Halloween IX: It Ain't Necessarily SCO

Rob Landley

Eric Raymond

Open Source Initiative
Revision History
Revision 1.32003-08-28rob
More updates from Craig Trader, Ed Hynan, Tony Taylor, Lon Jones, Vsevolod Sipakov, Earl Barr, Mark Taylor, Robert J. Harley, Eric Lesage, Andries Brouwer, Dick Porter, and dozens of other people who found the same things a bit later. (And me digging up yet more fresh material, and fixing some of the fresh material I dug up in 1.2.)
Revision 1.22003-08-26rob
Updates from Lars Nygard, Leon Brooks, Behrens Matt, Alastair Mayer, and a bunch of material from David Wheeler. Plus me getting kicked in the butt by the above to polish and fill in some bits. (And typos, typos, typos...)
Revision 1.12003-08-26esr
First round of minor corrections from the net.
Revision 1.02003-08-24rob
First published version.
Revision 0.12003-06-28esr
Document begun.

Table of Contents

Editors' Introduction
Header of the Complaint
Nature of This Action
Parties, Jurisdiction and Venue
Background Facts
The UNIX Operating System
SCO's Creation of a Market for Intel - The Genesis of SCO OpenServer
The SCO OpenServer Libraries
SCO's Development of UnixWare on Intel
Project Monterey
The AT&T UNIX Agreements
The IBM Related Agreements
The Sequent Agreements
Marketplace Value of UNIX
The Functional Limitations of Linux Before IBM's Involvement
IBM's Scheme
IBM's Coordination of Linux Development Efforts
(Breach of IBM Software Agreement)
(Breach of IBM Sublicensing Agreement)
(Breach of Sequent Software Agreement)
(Unfair Competition)
(Interference with Contract)
(Misappropriation of Trade Secrets—Utah Code Ann. §13-24-1 et seq.)
Prayer for Relief
Jury Trial Demand
A. The Decline and Fall of Caldera
B. Sizing the Unix Market
C. Moore's Law
D. The Itanic Disaster
Why Itanium?
Why Itanium Failed
The Monterey Fiasco

Editors' Introduction

The amended SCO complaint against IBM filed on 16 June 2003 is, like its predecessor, a tissue of lies, deliberate distortions, and flimflam. This complaint strongly suggests that SCO has no real case, since it contains so many false and misleading statements. Instead, it appears that SCO is trying to turn worthless products into money by running up the stock price on false claims and/or trying to get bought out so the frivolous legal actions will stop.

Unlike its predecessor, this amended complaint has been brought to you by the generosity of Microsoft, who (on the evidence of SCO's 10-Q SEC filings) dropped at least six million dollars on SCO (plus a promise of five million more over the next three quarters) to help it make trouble for Linux.

SCO, having willingly made itself a sock puppet for the boys in Redmond, therefore becomes the first company other than Microsoft to have its utterances admitted to the gallery of infamy that is the Halloween Documents.

There follows the usual point-by-point takedown. Unlike SCO's claims, this analysis is based entirely on public information which third parties may verify by chasing links or through their local library.

Text in fixed-width font is verbatim from the complaint. Where our commentary is interspersed with material from the original complaint, our commentary is indented and highlighted in green as well. Key words and particularly egregious falsehoods in the original are highlighted in red.

We've also added some appendixes at the end to explain the history of this lawsuit: The Decline and Fall of Caldera, Sizing the Unix Market, Moore's Law (and how it applies here), and The Itanic Disaster.

Header of the Complaint


Brent O. Hatch (5715)
Mark F. James (5295)
10 West Broadway, Suite 400
Salt Lake City, Utah  84101
Telephone:  (801) 363-6363
Facsimile:  (801) 363-6666

David Boies
333 Main Street
Armonk, New York 10504
Telephone:  (914) 749-8200
Facsimile:  (914) 749-8300

Stephen N. Zack
Mark J. Heise
100 Southeast Second Street
Suite 2800
Miami, Florida  33131
Telephone:  (305) 539-8400
Facsimile:  (305) 539-1307

Attorneys for Plaintiff The SCO Group, Inc.

                             DISTRICT OF UTAH

|                                      |                                      |
| THE SCO GROUP, INC.,                 |                                      |
| a Delaware corporation,              |                                      |
|                                      |       AMENDED COMPLAINT              |
|       Plaintiff,                     |                                      |
| vs.                                  |        (Jury Trial Demanded)         |
|                                      |                                      |
| INTERNATIONAL BUSINESS MACHINES      |                                      |
| CORPORATION, a New York corporation, |        Case No. 03-CV-0294           |
|                                      |                                      |
| Defendant.                           |        Hon: Dale A. Kimball          |


Plaintiff, The SCO Group, Inc., a Delaware corporation, f/k/a Caldera Systems, Inc. (“SCO”), sues Defendant International Business Machines Corporation (“IBM”) and alleges as follows:

The plaintiff in this case, "SCO Group Inc.", was originally the Linux vendor "Caldera Systems Inc." It changed its name to "Caldera International Inc." on October 11, 2000 after purchasing the server division of The Santa Cruz Operation, and changed its name again to "SCO Group Inc." on May 15, 2003. The original SCO changed its name to "Tarantella Inc" on February 12, 2001 after divesting itself of its deteriorating Unix business to focus on its Tarantella application.

These are two distinct companies, one founded in 1979 and headquartered in California, and the other founded in 1994 and headquartered in Utah, which the plaintiff refers to interchangeably as SCO. To distinguish them, we refer to the plaintiff as SCO/Caldera and the older company the plaintiff purchased a server division from as SCO/Tarantella.

For the sad tale of Caldera's degeneration from an open-source pioneer to a parasitic lawsuit factory, see Appendix A.

Nature of This Action

1. UNIX is a computer operating system program and related software originally developed by AT&T Bell Laboratories (“AT&T”). UNIX is widely used in the corporate, or “enterprise,” computing environment.

Actually, ‘UNIX’ is not “a computer operating system program”. There is no single operating system you can point to and say "that alone is Unix, everything else is not Unix". Technically, the name "Unix" is a trademark of The Open Group describing a group of operating systems conforming to The Open Group's ‘Single Unix Specification’. More generally, Unix is a generic technical term used to describe a family of over a hundred operating systems, some of which meet The Open Group's trademark standard but many of which do not. Compaq's Tru64 was submitted and qualified for the Unix trademark, while Apple uses Unix in a generic sense when referring to its MacOS X. Microsoft's Windows NT has hosted a Unix compatibility environment called cygwin since 1998, allowing it to run Unix software.

While all Unixes in either sense share common ideas and methods, the ideas and methods they share are (and for twenty years have been) part of the common state of engineering art in the construction of operating systems, extensively documented in the public literature, and shared by a global community of hundreds of thousands of programmers and software engineers. The first paragraph of this complaint is the beginning of a deliberate, mendacious attempt to misrepresent this entire body of work as SCO's property.

SCO does not even own the Unix trademark, and they cannot be be ignorant of The Open Group, since both SCO and IBM are members. IBM is one of five "Sponsors", and SCO is one of 138 "Regular Members" — IBM's membership in The Open Group significantly outranks SCO's.

SCO itself does not offer a product named just ‘Unix’, it offers UnixWare (named after Novell Netware) and SCO OpenServer (which was once Microsoft Xenix).

Over the past 30 years, the basic ideas of Unix have permeated every aspect of the software industry. In 1996, Microsoft CEO Bill Gates gave a speech at Unix Expo where he called Windows NT "a form of Unix". The same "UNIX 95" certification SCO obtained for UnixWare was also granted to IBM's OS/390 mainframe operating system. A decade ago, IBM's OS/2 operating system for the PS/2 had an open source Unix emulation package called EMX (which was not created by IBM, but by a german individual named Eberhard Mattes).

Over the past 30 years, the basic ideas of Unix have permeated every aspect of the software industry. In 1996, Microsoft CEO Bill Gates gave a speech at Unix Expo where he called Windows NT "a form of Unix".

2. Through a series of corporate acquisitions, SCO presently owns all right, title and interest in and to UNIX and UnixWare operating system source code, software and sublicensing agreements, together with copyrights, additional licensing rights in and to UNIX and UnixWare, and claims against all parties breaching such agreements. Through agreements with UNIX vendors, SCO controls the right of all UNIX vendors to use and distribute UNIX. These restrictions on the use and distribution of UNIX are designed to protect the economic value of UNIX.

The “series of corporate acquisitions” draws a kindly veil over the fact that at least three different companies — AT&T, Novell, and the original SCO (now Tarantella) — have each discarded the rights to the increasingly obsolete System V codebase. Along the way, the intellectual property was significantly impaired.

The most obvious impairment of SCO's “right and title” to Unix source code and related IP is the 1993 settlement with the University of California at Berkeley, which legally confirmed the plurity of Unix by severing the BSD source base, despite its origin as a derivative work of AT&T code, from the control of AT&T's successors in interest.

USL's willingness to settle the lawsuit, "Unix System Laboratories v. Berkeley Software Design, Inc." (1993 U.S. Dist. LEXIS 19503; 27 U.S.P.Q.2D (BNA) 1721.), was motivated by the judge's ruling against issuing a preliminary injunction, which read in part:

Since Plaintiff has failed to provide enough evidence to establish a "reasonable probability" that Net2 or BSD/386 contain trade secrets, I find that Plaintiff has failed to demonstrate a likelihood of success on the merits of its claim for misappropriation of trade secrets. No preliminary injunction will issue.

The settlement resulted in the release of 4.3-Lite BSD, an open-source version of Unix explicitly free of any intellectual property claims from AT&T or its successors. The existence of unimpaired BSD imposed a massive new burden on any intellectual property claims putatively based on ownership of the System V code: to prove that any disputed IP could not have come from BSD.

Another significant source of impairment is the plethora of existing "perpetual" Unix licenses. These licenses were granted in return for a one-time payment requiring no ongoing royalties; they never expire, and they allow unlimited sublicensing and the creation of derivative works that explicitly belong to the licensee. IBM's original license to System V technology from AT&T is such a license, as shown by language in "Exhibit C" on SCO's own website, which is an amendment/clarification of IBM's System V contract. Paragraph 2 of Exhibit C explicitly allows IBM to create derivative works, and affirms that those derivative works belong entirely to IBM (except for any actual System V code snippets included in them, if any).

Regarding Section 2.01, we agree that modifications and derivative works prepared by or for you are owned by you. However, ownership of any portion or portions of SOFTWARE PRODUCTS included in any such modification or derivative work remains with us.

Although IBM's perpetual System V code license was with AT&T, SCO is still selling these perpetual licenses as the core offering of their SCOsource initiative. The second full paragraph on page 28 of their 2003 first quarter SEC filing begins:

Our SCOsource licensing revenue to date has been generated from license agreements that are non-exclusive, perpetual, royalty-free, paid up licenses to utilize our UNIX source code, including the right to sublicense.

A third source of impairment is the simple fact that tens of thousands of copies of Unix source code have been circulating within the community for three decades. SCO itself puts the number at over 30,000 licensing agreements with 6000 entities (from the same 2003 Q1 SEC filing, bottom of page 20). Maintaining trade secret status on something like that is effectively impossible. One of the authors of this commentary is collecting evidence from senior Unix gurus who have source code to System V and derivatives in their archives, to prove that this code cannot have trade secret status.

It is worth noting that when SCO says “These restrictions on the use and distribution of UNIX are designed to protect the economic value of UNIX.” it is using the phrase “economic value” in a tendentious way. No action of IBM has altered the economic value of Unix; a copy of Unix has much the same effect on a user's ability to process data and do things as it did before any of IBM's actions. The price that can be charged for that value, the amount of rent SCO can collect, has been diminishing for years, but this is nothing new. SCO's own SEC filings have proposed a number of alternative hypotheses for their poor economic performance, and before 2003 SCO/Caldera had never had a single profitable quarter. They're not complaining about the loss of actual revenue, but about the loss of a purely hypothetical market opportunity that might conceivably have resulted in revenue.

3. A variant or clone of UNIX currently exists in the computer marketplace called “Linux.” Linux is, in material part, based upon UNIX source code and methods, particularly as related to enterprise computing methods found in Linux 2.4.x releases and the current development kernel, Linux 2.5.x. Significantly, Linux is distributed without a licensing fee and without proprietary rights of ownership or confidentiality.

Referring to Linux as a “clone” presumes the conclusion SCO wishes the court to reach. Linux is not in essence a clone or derivative work of any SCO property, but an autonomous line of evolution of generic, publicly documented Unix ideas, with a history of independent development documented back to its beginnings. Even a showing that Linux incidentally or accidentally contained some Unix code would not alter Linux's essential nature as an independent work.

Linus Torvalds explained the situation in the chapter he wrote for the 1999 book "Open Sources" (the complete text of which is available online):

Linux is a Unix-like operating system, but not a version of Unix. This gives Linux a different heritage than, for example, Free BSD. What I mean is this: the creators of Free BSD started with the source code to Berkeley Unix, and their kernel is directly descended from that source code. So Free BSD is a version of Unix; it's in the Unix family tree. Linux, on the other hand, aims to provide an interface that is compatible with Unix, but the kernel was written from scratch, without reference to Unix source code. So Linux itself is not a port of Unix. It's a new operating system.

A quick summary of early Linux development circa 1991 and 1992 is available from Linux International. (Notice that the first documentary evidence of the Linux project on the internet was Linus Torvalds asking for a copy of the Posix standard, not for source code.) The Linux development process has always been public; an archive of the "linux-activists" mailing list dating back to october 1991 is available online. The book "Just for Fun", by Linus Torvalds and David Diamond, covers the early days of Linux in detail. Linus created Linux shortly after taking a course on operating system design from the University of Helsinki, which used the textbook Operating Systems: Design and Implementation by Dr. Andrew Tanenbaum, which was about a tiny unix-like operating system called "Minix" which Tanenbaum had written from scratch as a teaching tool. [1] Linus chose a very different design from Minix, and used no Minix code in Linux, even though Minix itself was an independent work of Tanenbaum's which contained no System V code. As Linux became more popular, Tanenbaum publicly disagreed with Linus' design choices, and a subsequent public debate took place on the internet in January 1992 between Tanenbaum and Torvalds on operating system design principles. The complete source code of many Linux kernel versions dating back to 1991 is freely downloadable from www.kernel.org; you can download the very first release (or view the release notes), the first semi-stable version (release notes), or the 1.0 release. Linus Torvalds has been interviewed numerous times, here are a 1992 interview and a 1994 interview. Here is an academic analysis of the history of Linux kernel development. Here is a recent weekly technical summary of Linux-kernel development activity, which should give even a layman an example of the openly documented nature of the development, and may be considered a representative example from the archive of such technical summaries going back to the beginning of 1999.

Historian Peter Salus, author of "A Quarter Century of Unix" and executive director of Usenix for three years back in the 1980s, gave a talk about the History of Linux for the tenth anniversary of Linux, a recording of which is available online and as a downloadable MP3 (1 hour and 11 minutes). (He also gave an excellent talk at Usenix in 2001 on A History of Unix and Unix Licenses (50 minutes).)

Throughout this complaint, SCO attempts to confuse the widely admitted sharing of ideas and methods among computer professionals and Unix-family operating systems with putatively illicit misappropriation of source code. This attempt begins here with the introduction of the term “methods” But the history is clear; Linux could be “based on” Unix methods through published treatment such as Maurice Bach's The Design of the Unix Operating System, through the Berkeley Unix line of source code (also well documented in academic textbooks such as The Design and Implementation of the 4.4 BSD Operating System), or in several other ways that would not imply any legal connection to or proprietary entitlement by AT&T's successors in interest.

During the months since its initial complaint was filed, SCO has repeatedly refused to make public any actual evidence that Linux is “in material part, based upon UNIX source code”, preferring instead to trumpet its claims in the media. (Their one gesture in the direction of disclosure, at SCOforum in August 2003, proved to be of code they don't own.) In any case, given that the revision history of the SCO source code is not (unlike that of Linux) a matter of public record readily accessible on numerous archives, the burden should be on SCO to show that any similarities have not been covertly and deliberately inserted by SCO itself for the purpose of enabling SCO to commit barratry against IBM and the Linux community at large.

4. The UNIX software distribution vendors, such as IBM, are contractually and legally prohibited from giving away or disclosing proprietary UNIX source code and methods for external business purposes, such as contributions to the Linux community or otherwise using UNIX for the benefit of others. This prohibition extends to derivative work products that are modifications of, or based on, UNIX System V source code or technology. IBM and certain other UNIX software distributors are violating this prohibition, en masse, as though no prohibition or proprietary restrictions exist at all with respect to the UNIX technology. As a result of IBM's wholesale disregard of its contractual and legal obligations to SCO, Linux 2.4.x and the development Linux kernel, 2.5.x, are filled with UNIX source code, derivative works and methods. As such, Linux 2.4.x and Linux 2.5.x are unauthorized derivatives of UNIX System V.

This paragraph begins with a sleight of hand. By using the phrase “proprietary Unix source code and methods” in a way that leaves it ambiguous whether “proprietary” is a restrictive modifier or not, SCO attempts to confuse a trivial but true claim (disclosure of proprietary methods is forbidden) with an extremely broad and false claim (disclosure of Unix methods is forbidden).

But, in fact, it has not been shown that there are any proprietary methods or any proprietary source code at issue. SCO holds no Unix patents; the state and disposition of the Unix copyrights is unclear and presently disputed between SCO and Novell; finally, by making the source code “readily ascertainable by proper means” including the sale of source licenses to educational institutions, SCO and various predecessors in interest have failed to meet the Utah standards for maintenance of trade secrecy in the Unix source code. Their "trade secrets" have been taught in college classes for 20 years, with their knowledge and approval.

On August 18 2003 at the annual SCOforum show, SCO tried to back up its contention that Linux is “filled with UNIX source code, derivative works and methods” — and was promptly shot down.

As for derivative works, SCO needs to read its own Exhibit C, paragraph 2, which states “...modifications and derivative works prepared by or for you are owned by you.”. (The “you” in this case being IBM.) More on this in the response to paragraph 6.

As a final point, the moment that SCO shipped a Linux kernel containing any included System V code, it knowingly put that code under the terms of the GNU General Public License, and that code thereby ceased to be proprietary (and if SCO/Caldera did not accept the GPL, then they willfully violated the copyrights of every single Linux contributor). SCO and predecessor in interest Caldera shipped Linux source code for eight years, and joined numerous Linux development consortia including The Trillian Project and United Linux. Although SCO tried to cover its tracks by suspending distribution of its own Linux product two months after filing its lawsuit against IBM, Linux source code is still available for download from SCO, and must remain so for three years according to the terms of secton 3 of the GNU General Public License under which Linux was made available to SCO/Caldera for distribution. SCO/Caldera complied with its obligations under the GPL for eight years, and can hardly start claiming it was unaware of the terms of that license, since they submitted a complete copy of the GPL to the securities and exchange commission (attached as exhibit 10.14 to their S-1 Inital Public Offering filing [2]). The GPL was also the contract under which they sold Linux to their customers, thereby binding themselves to an obligation to those customers.

SCO voluntarily chose to participate in the Linux market. It was never required by any third party to do so, and found doing so profitable (in gross terms, if not net). If SCO didn't exert due diligence on the code SCO itself shipped, it can't blame IBM. In fact, Linux was SCO/Caldera's core business before it purchased the archaic Unix line from SCO/Tarantella three years ago. When Caldera first acquired SCO, its SEC filings stated the intention to “unify” the Linux and Unix lines, specifically questions 29 and 31 of the question and answer session:

29. How will the products of both companies be integrated?

We will integrate all products as appropriate based on Caldera, Inc.'s focus of common application, development and management.

31. How will this affect the Linux industry/community?

Caldera, Inc. intends to unify the technologies of Linux and UNIX enhancing the attractiveness to Linux and business customers. In addition, Caldera, Inc. will provide open access to those technologies, greatly benefiting the Linux community and industry.

Finally, this paragraph concludes with an attempt to parlay a narrow claim of IBM contract violations into a blatant land grab against all current versions of Linux, including millions of lines of code which are not derived from System V code by any standard, however inclusive. This is overreaching of the crudest sort.

5. As set forth in more detail below, IBM has breached its obligations to SCO, induced and encouraged others to breach their obligations to SCO, interfered with SCO's business, and engaged in unfair competition with SCO, including by:

a) misusing UNIX software licensed by SCO to IBM and Sequent;

b) inducing, encouraging, and enabling others to misuse and misappropriate SCO's proprietary software; and

c) incorporating (and inducing, encouraging, and enabling others to incorporate) SCO's proprietary software into Linux open-source software offerings.

In this summary of claims SCO speaks only of software, not of methods or algorithms. The implied breach is of SCO's copyrights (if any). Elsewhere, however, SCO speaks of methods. The effect is to deliberately confuse two issues and prejudice claims which SCO could not otherwise establish.

6. As a result of these breaches, SCO sent a notice of termination to Mr. Sam Palmisano, the Chief Executive Officer of IBM on March 6, 2003. The termination notice specified that, pursuant to SCO's contractual rights under controlling agreements, IBM's right to use or distribute any software product based on UNIX System V technology, including its own version of UNIX known as “AIX,” would be terminated on June 13, 2003, unless such breaches were reasonably cured prior to that time.

IBM has pointed to the plain language of its license agreement as a refutation of any claim of breach. This language, from Exhibit C, includes the following in paragraph 9:

Nothing in this agreement shall prevent LICENSEE from developing or marketing products or services employing ideas, concepts, know-how, or techniques relating to data processing embodied in SOFTWARE PRODUCTS subject to this agreement, provided that LICENSEE shall not copy any code from such SOFTWARE PRODUCTS into any such product [...] If information relating to a SOFTWARE PRODUCT subject to this agreement at any time becomes available without restriction to the general public by acts not attributable to LICENSEE or its employees, LICENSEE's obligations under this section shall not apply to such information after such time.—

Since SCO itself ensured that any System V code in Linux would become available without restriction to the general public by shipping it in a Linux kernel for eight years (up through version 2.4.14), the conditions of this language have been met. In addition, the source code for Unix System III, an earlier version of Unix containing a significant subset of the later System V code, was made available for free download on SCO's website under a click-through "Ancient Products License". And the BSD lawsuit confirmed the release of a significant portion of the methods, if not the actual code, of System V Unix.

7. The termination notice was based, in part, on IBM's self-proclaimed contributions of AIX source code to Linux, and use of UNIX/AIX methods for accelerating the development of Linux in contravention of IBM's contractual obligations to SCO.

IBM's license entitles it to use Unix concepts and methods, to make derivative works, and to sublicense Unix source code and technologies to others.

If SCO is going to try to quote IBM's marketing people out of context, the obvious response is to quote SCO's CEO in context. The CEO of Caldera during the SCO acquisition and for a number of years afterwards was Ransom Love. An interview Ransom Love gave in August 2001 (a year after the SCO acquisition) includes several interesting statements. In answer to the question “What is your position [on open source] now?”, Love replied “Caldera is absolutely committed to open source... We are going to continue publishing under open source.” In answer to the question "What does the future hold for your unified Linux/Unix platform?", Love replied "With UnixWare we can now take Linux to 32-way systems... So we are taking both and combining them into one platform." When asked about OpenServer, Love said "...what we plan to do is to take the OpenServer technology to Linux, probably with some sort of open source source licence..." When asked "What does the recent acquisition of SCO really mean for Caldera?", Love replied "Our mission is to enable development, deployment and management of a unified Linux and Unix operating system."

If the current management of SCO is completely unaware of what their CEO claimed the company's mission was two years ago, they are incompetent. If they are aware of it, they're being blatantly dishonest. SCO/Caldera devoted entire securities and exchange commission filings to unifying Linux and Unix.

8. Pursuant to its rights under the controlling agreements, IBM was entitled to 100 days to cure its underlying contractual breaches, provided it was willing and able to do so. Both parties were contractually required to “exert their mutual good faith best efforts to resolve any alleged breach short of termination.”

Nowhere does SCO specify exactly what actions SCO wanted IBM to undertake. What, specifically, did SCO want IBM to do? It can't have wanted IBM to remove offending code from the Linux Kernel, because it refuses to reveal to IBM or the Linux community which code it considers to be infringing. SCO CEO Darl McBride explained his reason for refusing to identify allegedly infringing code in the Linux kernel as follows: "The Linux community would have me publish it now, (so they can have it) laundered by the time we can get to a court hearing. That's not the way we're going to go." SCO actively does not want IBM or other Linux participants to stop infringing, and has intentionally made it impossible for them to comply with SCO's stated wishes. This cannot be considered a good faith best effort on SCO's part.

9. To that end, SCO did everything reasonably in its power to exert a good faith effort to resolve the termination of IBM's UNIX contract rights. Conversely, during the 100-day period, IBM did not set forth a single proposal or idea for cure.

SCO's phrasing here is particularly misleading. Notice they don't say they made any effort to avoid the termination of IBM's contract. In this context, “resolve” means “enact”, or “hasten”. The contract language (again in Exhibit C, paragraph 5 this time) forced them to wait 100 days. It's too bad that SCO's promise not to terminate for “breaches we consider to be immaterial” didn't require arbitration by a disinterested third party. Asking SCO to judge what is and isn't immaterial is like asking the fox to guard your chickens.

As for the rest of the paragraph, IBM has maintained that there is no breach. Where there is no breach, there is no need for cure.

10. SCO has therefore terminated IBM's right to use any part of the UNIX System V source code, including its derivative AIX, effective as of June 13, 2003 (the “AIX Termination Date”).

11. As of the AIX Termination Date, IBM is contractually obligated to discontinue use of and return or destroy any and all copies of the Software Products defined in the controlling agreements, which include UNIX System V source code and all its derivatives, including AIX.

As a matter of public record, IBM has maintained that its license is perpetual and irrevocable. This position is based on the language of Amendment X, dated 1996.

It's also interesting to note that that the “perpetual” licenses that SCO is offering through SCOsource are the same general kind of license: that is, they seem only to be guaranteed to last until SCO needs money.

Parties, Jurisdiction and Venue

12. Plaintiff SCO is a Delaware corporation with its principal place of business in Utah County, State of Utah.

13. Defendant IBM is a New York corporation with its principal place of business in the State of New York.

This case was removed from Utah state to Federal court because of errors in SCO's original complaint. One of these was giving IBM's place of business incorrectly. Nice to see they got it right this time.

14. Sequent Computer Systems, Inc. (“Sequent”) was formerly an Oregon corporation that contracted with SCO's predecessor in interest, AT&T. Sequent was subsequently merged into IBM in a stock transaction.

Sequent started in 1983 when a group of engineers left Intel to work on parallel computing. Their first product, released in 1984, ran a modified version of BSD called Dynix. Their first Intel 80386 based product shipped commercially in 1987. Like many in the industry, Sequent shifted from BSD to System V in response to the uncertain legal status of the BSD code before the USL vs BSD lawsuit was resolved. The (theoretically) System V based product was called Dynix/ptx. In reality, Dynix/ptx was still based on BSD, with some System V interface wrappers so it could claim System V compatability, and an AT&T license because the USL vs Berkeley lawsuit hadn't been resolved in Berkeley's favor yet. (Wikipedia has an excellent article on Sequent's history.)

15. This Court has subject matter jurisdiction pursuant to 28 U.S.C. §1332 in that diversity of citizenship exists between the parties and the amount in controversy exceeds $75,000, exclusive of interest and costs.

16. This Court has in personam jurisdiction over IBM pursuant to Utah Code Ann. §78-27-24 on the bases that IBM is (a) transacting business within this State, (b) contracting to provide goods and services within this State and (c) causing tortious injury and breach of contract within this State.

Point c. being the claim they hope to prove, not a statement of fact.

17. Venue is properly situated in this District pursuant to 28 U.S.C. §1391 in that IBM maintains a general business office in this District and a substantial part of the events giving rise to the claims alleged herein occurred in this District.

If SCO's claims had any merit, the latter assertion would almost certainly be false. IBM's largest Linux development center in the continental US is the Linux Technology Center in Austin, Texas. SCO's corporate headquarters during the Monterey project (the period of the disputed contract) was in California, with a large developer concentration in Canada. Since the "events giving rise to the claims alleged herein" occurred largely in the fevered imagination of various SCO executives and nowhere else, a case could be made that that they took place in Utah. The reality is that Utah is the venue only because SCO wanted a hometown advantage, and it's dishonest to claim otherwise. (This is a trivial error in comparison to those in the rest of their complaint.)

Background Facts

The UNIX Operating System

18. UNIX is a computer software operating system. Operating systems serve as the link between computer hardware and the various software programs (“applications”) that run on the computer. Operating systems allow multiple software programs to run at the same time and generally function as a “traffic control” system for the different software programs that run on a computer.

This claim is technically defective. Not all operating systems allow multiple programs to run at the same time.

19. By way of example, in the personal computing market, Microsoft Windows is the best-known operating system. The Windows operating system was designed to operate on computer processors (“chips”) built by Intel. Thus, Windows serves as the link between Intel-based processors and the various software applications that run on personal computers.

20. In the business computing environment for the Fortune 1000 and other large corporations (often called the “enterprise” environment), UNIX is widely used.

However, SCO's own versions of Unix have never had significant enterprise market share. The technology to support that use had to be supplied by others, as we document in the first OSI Position Paper on the SCO-vs.-IBM Complaint.

In terms of revenue, the top three players in 1999 were Sun (38 percent of the market), Hewlett Packard (24 percent), and IBM (16 percent), together soaking up a combined total of 78 percent of the Unix market's revenue and leaving SCO/Tarantella part of a long list of financial also-rans. SCO's market share came by pricing their systems so low they wound up in financial difficulty and sold their Unix business to Caldera, to finance a transition to something they considered potentially more profitable.

To answer the question of exactly how widely SCO's own Unix products are used, it's necessary to point out that SCO sells two very different versions of Unix. SCO OpenServer is a low-end product that traces its heritage back to Microsoft Xenix in 1979. OpenServer is primarily used for point of sale systems such as fast food cash registers. SCO's other Unix line is UnixWare, which was named after Novell NetWare and is the Unix product SCO acquired from Novell in 1995. SCO put three years of intense development effort into UnixWare between 1995 and 1998 in an attempt to make an enterprise operating system out of it, but never managed to achieve any significant marketshare or market interest.

SCO's low-end product, OpenServer, provides the majority of their product revenue and the vast majority of their unit volume. According to page 22 of SCO's first quarter 2003 SEC filing::

SCO's first quarter 2003 Company's products revenue.

For the second quarter of fiscal year 2003, OpenServer revenue was $5,750,000, or 52 percent of total products revenue and UnixWare revenue was $3,017,000, or 27 percent of total products revenue. For the first two quarters of fiscal year 2003, OpenServer revenue was $10,491,000, or 47 percent of total products revenue and UnixWare revenue was $7,181,000, or 32 percent of total products revenue.

SCO is not a significant "enterprise Unix" player not only because their entire quarterly revenue from their "enterprise" product (UnixWare) is less than 1/10th that of Sun, IBM, or HP (which is not a new development), but because their low-end product provides the majority of their products revenue, and only on a single hardware platform. They are primarily a niche player in the low-end PC Unix market.

It's also interesting to note that any special heritage from AT&T that SCO claims to have acquired through Novell would have been embodied in UnixWare, yet UnixWare is not SCO's best-selling product. And OpenServer, which provides the majority of their revenue today, was at one point a discontinued piece of trash that Caldera wanted to leave with SCO/Tarantella when it first tried to purchase SCO's operating systems division, as explained in the February 9, 2001 article, "Caldera Systems expands Unix acquisition plans":

Under the previous terms of the acquisition, announced in August, Linux seller Caldera Systems planned to acquire SCO's UnixWare product. Now it plans to also acquire SCO's earlier version of Unix, OpenServer, Caldera said in a statement Friday.

SCO stopped development of OpenServer in 1999, and Caldera Systems plans to continue with SCO's practice of only minor updates while encouraging customers to migrate to other operating systems, Tamang said.

OpenServer remains SCO's cash cow (in gross, if not net terms) because they were unsuccessful in their efforts to migrate customers off of it, just as they were unsuccessful in their efforts to repurpose the acquired reseller network to sell Linux instead of obsolete proprietary versions of Unix.

21. The UNIX operating system was originally built by Dennis Ritchie, Ken Thompson and other software engineers at AT&T Bell Laboratories. After successful in-house use of the UNIX software, AT&T began to license UNIX as a commercial product for use in enterprise applications by other large companies.

This paragraph ignores huge contributions by many developers outside of AT&T, many of whom are active in the Linux community today. Specifically, at the time Unix was invented at Bell Labs, AT&T was forbidden from commercializing it by a Jan 24, 1956 antitrust consent decree[3]. Specifically, "An injunction was issued which barred AT&T from engaging in any business other than the provision of common carrier communication services, and required Western Electric and AT&T to license their patents to anyone who wanted them upon the payment of appropriate royalties."

Since AT&T was forbidden from entering the computer business, they could not sell Unix as a product. The outside world found out about Unix after Ken and Dennis presented a paper on their Unix research at a 1973 ACM operating systems conference, and the paper was widely published to the academic community in the proceedings of that conference in 1974. The 1956 consent decree required AT&T to license its technology to outsiders who asked, but their offical policy concerning these academic Unix licensees was: "No advertising, no support, no bug fixes, payment in advance."

Unix could not become a commercial product of AT&T before the 1984 divestiture undid the 1956 antitrust injunction keeping them out of the computer business. During this decade-long gap, there was no commercial incentive for AT&T to invest in Unix development, but significant incentive for early open-source development to occur, largely co-ordinated by the University of California at Berkeley after Ken Thompson's teaching stint there (see the response to Paragraph 3).

Ignoring these contributors is crucial to SCO's case; were it to acknowledge them, it would have to admit to the substantial impairment of its rights by the settlement of the 1993 BSD case, which hinged on AT&T's misappropriation of work done at Berkeley and elsewhere. See, again, the OSI Position Paper on the SCO-vs.-IBM Complaint.

22. Over the years, AT&T Technologies, Inc., a wholly owned subsidiary of AT&T, and its related companies licensed UNIX for widespread enterprise use. IBM, Hewlett-Packard, Inc. (“HP”), Sun Microsystems, Inc. (“Sun”), Silicon Graphics, Inc. (“SGI”) and Sequent became some of the principal United States-based UNIX licensees, among many others.

It is noteworthy that they do not mention themselves, the original Santa Cruz Operation, SCO/Caldera's predecessor in interest. SCO sublicensed Xenix from Microsoft, and in 1987 Microsoft divested itself of Xenix to focus on DOS, OS/2, and Windows. Xenix went on to become SCO's primary product, and remains so today under the name SCO OpenServer. In the 1980s, SCO was one of many licensees of AT&T System V Unix. Perhaps they belong in the "among many others" category because they were not an enterprise player.

Microsoft's CEO Bill Gates reminisced about Microsoft Xenix in a 1996 speech at Unix Expo, where he was trying to recruit Unix developers to port their applications to Windows NT:

If we go way back in time, Microsoft was actually the first one to go to AT&T and beg to get a nice high-volume commercial license for Unix. And for many, many years we were the highest volume licensee, not only for our own Xenix products, but Siemens with theirs, Santa Cruz with theirs, and dozens and dozens of sub-licensees.

I have to admit, it was fairly difficult to work with AT&T back then. They simply didn't understand what they had. They didn't understand how to manage the asset, either in terms of promoting it properly or in terms of making sure that there wasn't fragmentation in how different implementations were put together. And so that vacuum in leadership created a bit of a dilemma for everybody who was involved in Unix.

Well, Microsoft stepped back and looked at that situation and said that the best thing for us might be to start from scratch: build a new system, focus on having a lot of the great things about Unix, a lot of the great things about Windows, and also being a file-sharing server that would have the same kind of performance that, up until that point, had been unique to Novell's Netware.

And through Windows NT, you can see it throughout the design. In a weak sense, it is a form of Unix. There are so many of the design decisions that have been influenced by that environment. And that's no accident. I mean, we knew that Unix operability would be very important and we knew that the largest body of programmers that we'd want to draw on in building Windows NT applications would certainly come from the Unix base.

23. IBM, HP, Sun, SGI and the other major UNIX vendors each modified UNIX to operate on their own processors. Thus, for example, the operating system known as “HP-UX” is HP's version of UNIX. HP-UX is a modification of and derivative work based on UNIX System V source code.

And Microsoft modified Unix to operate on cheap 16 bit 8086 PCs which were primarily used for running DOS, and sublicensed it to the Santa Cruz Operation.

24. Similarly, the operating system known as Solaris is Sun's version of UNIX. Solaris is a modification of, and derivative work based on, UNIX System V source code.

Solaris is the second major Unix operating system from Sun. Its first, SunOS, was based on BSD code rather than AT&T's version, as the founding of Sun predated the AT&T divestiture and one of Sun's four founders, Bill Joy, had been Berkeley's head of Unix development as a graduate student. After the divestiture, AT&T purchased 19 percent of Sun and convinced them to migrate from BSD to the System V codebase (hence from SunOS to Solaris). The real picture is far more complicated than SCO presents: Commercial Unix existed before AT&T commercialized its own Unix. When AT&T entered the Unix market after the divestiture, it found itself in competition with the existing Unix players, leading to "the Unix wars" and the spin-off of Unix System Labs (USL) from AT&T (and Novell's subsequent purchase of USL).

25. SGI's UNIX-based operating system is known as “IRIX.” IRIX is a modification of, and derivative work based on, UNIX System V source code.

26. IBM's UNIX-based operating system is known as “AIX.” AIX is a modification of, and derivative work based on, UNIX System V source code.

Again, SCO oversimplifies. According to page 26 of the book "Life with Unix" by Don Libes and Sandy Ressler, IBM entered the Unix market in 1983 with a version of Unix (for the 16-bit 8086 based IBM PC) sublicensed from a company called “Interactive”. This predated AIX by a number of years. In the late 1980s, IBM also offered another version of Unix as an alternative to its System V based AIX: the BSD-based Academic Operating System (AOS). AIX is still around while IBM's other Unix offerings have died out because over the years IBM invested significant amounts of internal development effort into keeping AIX relevant.

27. Sequent's UNIX-based operating system is known as “DYNIX/ptx.” DYNIX/ptx is a modification of, and derivative work based on, UNIX System V source code.

See the response to paragraph 14 for the history of Dynix.

28. The various identified versions of UNIX are sometimes referred to as UNIX “flavors.” All commercial UNIX “flavors” in use today are modifications of and derivative works based on the UNIX System V Technology (“System V Technology”). Were it not for UNIX System V, there would be no UNIX technology or derivative works available for IBM and others to copy into Linux.

This last is a serious and willful misrepresentation of the facts. It is historically true that many (though not all) modern Unixes derive methods and ideas from System V. But the 1993 settlement, which forbade Novell and all their successors in interest (including SCO) from asserting further IP claims against the Berkeley source code, established that mere derivation of methods is not sufficient to establish a proprietary claim on the code.

SCO's claim that no commercial Unix product exists independent of System V means that BSDi's successor in interest, Wind River, is also a target despite the 1993 settlement. [4]

SCO also deliberately ignores other lines of Unix development, like SunOS, which either predated System V or were based on BSD. The claim in paragraph 28 is simply false. It is factually untrue. Sun was quite profitably selling SunOS from its founding until AT&T purchased a significant equity stake in the company to convince it to change horses. IBM could easily have made a commercial success of AOS. The Santa Cruz Operation's founding in 1979, and the early history of its core product OpenServer/Xenix, predates the AT&T divestiture and corresponding release of System V as the first commercial software product from AT&T. Again, SCO wilfully ignores the impact of a decade of Berkeley led open-source development of BSD, because the truth undermines their case.

29. SCO is the sole and exclusive owner of all Software and Sublicensing Agreements that control use, distribution and sublicensing of UNIX System V and all modifications thereof and derivative works based thereon. SCO is also the sole and exclusive owner of copyrights related to UNIX System V source code and documentation and peripheral code and systems related thereto.

Both these claims are materially false for reasons already adduced, although this is the first time they mention copyright. Again, they are attempting to confuse issues; they are not actually making copyright claims.

The copyrights to System V may also have been impaired by the BSD lawsuit, the issue was left unresolved and would have to be reopened if System V copyright claims were to be pressed. AT&T included code from BSD, SunOS, Xenix, and possibly other sources in System V, but did not always get proper permission, and even stripped off third-party copyright notices from the code it incorporated. It assumed that its ownership of Unix gave it carte blanche ownership of derivative works, an assumption which came back to haunt it during the BSD lawsuit. The one thing SCO probably fears most is a judge taking the 1993 settlement as a precedent.

In addition, Novell, the company that owned System V before SCO/Tarantella, has stated claims that the copyrights did not pass to SCO but remain with Novell. SCO/Caldera claims to have come up with an amendment to the contract between SCO/Tarantella and Novell, which Novell itself cannot find a copy of in its files. This issue remains unresolved, and Novell continues to strongly object to SCO's complaints against IBM and the Linux community.

Even if SCO/Caldera's amendment proves to be genuine, SCO had not registered a transfer of copyright with the Patent and Trademark office at the time of the alleged contract violation, or even at the time of filing its first complaint against IBM in Utah court. As such, any damages SCO could potentially be awarded against IBM for violation of those unregistered copyrights are capped by statute at $150,000 per infringing work. Considering that Microsoft has already given SCO several million dollars to pursue the lawsuit against IBM and Linux, and they've already spent around two million dollars on legal fees, copyright damages are not significant.

30. During the 1990s the enterprise computing market for high-performance workstation computers came to be dominated by UNIX and the primary UNIX vendors identified above, each supplying its own version of the UNIX operating system based on UNIX System V Technology. UNIX became synonymous with “workstation” computers that typically operated on a RISC processing platform.

Sun became the leading vendor in the 1980s with SunOS, which was based on BSD. It switched to System V years later, motivated by a large equity investment from AT&T.

SCO/Tarantella was not a workstation vendor on a RISC platform. Until 1995 it exclusively sold low-end Unix systems for cheap PC hardware. Meanwhile SCO/Caldera was founded in 1994 as a commercial Linux vendor, also on cheap PC hardware, and had no involvement with System V before purchasing SCO's server division in early 2000.

Interestingly, AT&T directly sold "386 Unix System V" (in competition with SCO and other PC Unix vendors) in 1987, on 3.5" and 5.25" floppy disks. This was during AT&T's brief attempt to become a PC vendor (as documented in this May 1986 article from Byte Magazine reviewing one of their hardware offerings). AT&T was remarkably unsuccessful competing against both commodity PC hardware and the high-end workstation market led by established vendors like Sun, and quickly spun off Unix Labs as a seperate company, which was acquired by Novell.[5]

31. The RISC processing platform provides high-power computing capabilities at a relatively higher price for “workstation” computing. The alternative to “workstation” computing is commonly known as “desktop” computing on personal computers. The operating system market for “desktop” personal computers is dominated by Microsoft Corporation and its various Windows-based operating system products. The reason for this distinction is that most desktop computers (PC's) are designed to operate on Intel and Intel-compatible computing platforms. Most workstations are designed to operate on variants of RISC processing platforms and RISC-compatible computing platforms. PC systems and RISC systems are not compatible with each other. Thus, most versions of UNIX will not operate on Intel-based PC's for desktop computing; and Windows will not operate on RISC-based workstations for enterprise computing.

SCO glosses over the fact that IBM invented RISC technology. SCO also ignores the existence of PC ports of several Unix variants, such as IBM's AIX for the PS/2, or Sun's Solaris x86. In terms of Windows not operating on RISC systems, remember that Microsoft sold a version of Windows NT for the Alpha processor, a RISC system, for several years before deciding there wasn't enough money in it. (Versions of Windows NT 4.0, up until service pack 3, were also available for the PowerPC and Mips processors, also RISC systems.)

32. Most of the primary UNIX vendors identified above did not attempt to develop a UNIX “flavor” to operate on an Intel-based processor chip set. This is because the earlier Intel processors were considered to have inadequate processing power for use in the more demanding enterprise market applications.

Compare this language with that of the original complaint, in which SCO falsely claimed to have been the sole 386 vendor in this market. Of the vendors SCO has listed so far (IBM, Sun, HP, SGI, and Sequent), the majority did produce Intel versions. IBM had AIX for the PS/2, Sun still sells Solaris for x86 (and gave away copies for several years under its Free Solaris Binary License Program), and Sequent shipped its first 386 based NUMA system back in 1985. That's three of the five, which is "most".

SCO's Creation of a Market for Intel - The Genesis of SCO OpenServer

The very title of this section implies a fallacy. The IBM PC was created by IBM. Shipping millions of units of hardware annually, which were powered by Intel processors, created a market niche for software vendors like SCO.

Furthermore, SCO's first Unix product, Xenix, was commissioned by Microsoft, the company behind the dominant PC operating system "Windows" as described in Paragraph 19 of this complaint. Microsoft was the AT&T licensee and the owner of the Xenix copyrights until 1987. SCO sublicensed Xenix from Microsoft, and was at the time one of several Xenix resellers for Microsoft. If SCO is claiming that the products it shipped created a market for Unix on IBM PC hardware, the credit would belong to Microsoft, not SCO.[6]

Also, the title is inaccurate: calling the x86 market the "Intel" market is like calling the PC market the "IBM" market. Just as the IBM PC defined a commodity niche for many similar computer systems, so has Intel's x86 line spawned a range of clones. Companies other than Intel currently active in the x86 processor market include Advanced Micro Devices (makers of the AMD Athlon, and important for reasons explained in Appendix D), National Semiconductor (big in laptops and low power devices), and Transmeta (Linus Torvalds' former employer). The Unix operating system family itself forms a third commodity niche, along with commodity PC systems running commodity x86 processors.

33. As computers grew in popularity to perform business functions, the processing power of Intel-based processor chips also began to increase dramatically. Consistent with Intel founder Gordon Moore's prediction, computer chips remained inexpensive while exponentially increasing in power and performance.

SCO doesn't bother to explain Moore's Law, which is important for understanding both why there was no special insight in SCO/Tarantella's business plan and why the System V technology SCO touts so highly is nearly valueless today.

Appendix C describes Moore's Law in more detail, along with several consequences of Moore's Law that SCO would rather avoid mentioning. Among the points covered in the appendix are that Moore's law offered no special insight to SCO/Tarantella that was not available to other vendors, that Moore's Law was not limited to Intel chips but applies to competing products (such as Motorola's 68k processor line, the main competitor to Intel chips in the 1980's), and that an inescapable consequence of Moore's Law is an annual depreciation rate for computer technology that renders System V technology nearly valueless a decade and a half after its introduction.

34. Seeing this emerging trend, it became evident to SCO that Intel chips would gradually gain widespread acceptance for use in the enterprise marketplace.

It was clear to all that microchips would eventually take over the enterprise market, due to Moore's Law. But the x86 line's dominance among microchip vendors was not the result of Moore's Law, which affected them all roughly equally, but due to IBM's endorsement of one specific microprocessor line (Intel's x86) for its PC, instead of Motorola's 68k, Zilog's Z8000, or other competing microprocessor lines. If IBM had gone with another processor for its PC, its descendants are what modern computers would be using today. No reasonable industry expert would dispute this.

35. Therefore, while other major UNIX vendors modified UNIX for their own respective RISC-based computing platforms, SCO developed and licensed the UNIX-based operating system for Intel-based processors for enterprise use that is now known as “SCO OpenServer.”

Here SCO/Caldera is explicitly claiming that SCO OpenServer, the successor to Xenix, was intended for enterprise use. Not UnixWare, but OpenServer, which had its development discontinued in 1999 under SCO/Tarantella due to irredeemable obsolecense. They seem unclear on the definition of "enterprise", which may be one reason they've never been successful in the enterprise market. OpenServer is used to run cash registers, not Fortune 500 data centers.

Background on the economic impact of Moore's Law is provided in Appendix C.

36. SCO's early engineers faced difficult design challenges in modifying UNIX for effective use on an Intel processing platform. The principal design constraint centered around the limited processing power the Intel chip possessed in the early 1980's. The Intel chip (designed as it was for personal computers) was not nearly as powerful as the enterprise chips used by IBM, Sun, SGI and others in their respective UNIX offerings.

They admit that the early 16-bit Intel chips which Microsoft comissioned Xenix for more than 20 years ago were of "limited processing power". The Xenix system tailored to them was also, unsurprisingly, of "limited processing power".

Only the high price of memory kept Intel chips competitive for mainstream PC users in the 1980s. Since a single megabyte of memory cost over $5000 when the IBM PC was introduced (see the premier issue of PC Magazine, February 1982); a few hundred dollars for the processor was irrelevant at that point. A 32-bit processor's ability to deal with much larger amounts of memory could not be brought to bear until the price of memory came down (by which point Intel had added 32-bit processors to its line and most Unix vendors had begun experimenting with RISC processor designs). However, followers of Moore's Law knew that 32-bit capability would be necessary in future (due to memory prices dropping 50% every 18 months), which is why serious Unix vendors didn't want to waste time on 16-bit platforms. In the early 1980s, Moore's Law was actually an argument _against_ 16-bit Intel processors.

37. Based on the early design constraint of Intel's limited processing power, SCO found an appropriate enterprise market niche for the early versions of SCO OpenServer—single-purpose applications such as point-of-sale control, inventory control and transactions processing, with the highest possible reliability. Intel processors were fully capable of performing these relatively simple, repetitive tasks, and could do so at a lower cost and as reliably as the more powerful enterprise processing platforms sold by the other UNIX vendors, such as Sun and IBM.

The PC's default operating system, DOS, could run any of the "single purpose applications" they list, and it doesn't even multitask. Xenix's selling point was "lower cost" than other Unix systems, especially the ability to squeeze into a tiny amount of expensive memory. SCO seems to think "enterprise" is synonymous with "cheap, low-end market niche". They're proud their OS was capable of performing simple, repetitive tasks at a lower cost than Sun and IBM. To them, this means enterprise.

38. One example of a customer well suited to the earlier version of SCO OpenServer software is McDonald's Corp. McDonald's has thousands of stores worldwide and needs all stores to operate on an integrated computing platform for ease of use, immediate access to information and uniformity. However, the actual computing requirements for each individual McDonald's location are functionally simple—sales need to be tracked and recorded, and inventory functions need to be linked to sales. SCO OpenServer reliably fulfills McDonald's computing requirements at reduced cost.

Again, SCO considers running McDonald's cash registers to be "an appropriate enterprise market niche". This is a task which any modern laptop computer could run in the background without the user at the keyboard noticing. A battery-powered portable CD player has enough computing ability to run a McDonald's cash register. Either they're intentionally lying, or they honestly don't know what "enterprise" means in the operating systems market.

Numerous DOS-based point of sale systems are still for sale, such as apos, posey, Invoice Writer... SCO is basically claming that Xenix was competitive with 16-bit single tasking DOS. Repeating the word "enterprise" in this context renders it meaningless.

39. SCO's business model for SCO OpenServer provides enterprise customers the reliability, extensibility (ease of adding or changing functionality), scalability (ease of adding processors or servers to increase processing power) and security of UNIX—but on inexpensive Intel processor chips. This combination allowed customers to perform an extremely high number of transactions and, at the same time, gather and present the information from those transactions in an economical and useful way for enterprise decision makers.

SCO was not alone in the PC market. In the late 1980s and early 1990s, a number of other companies provided Unix versions and Unix clones for x86 PC systems, and any of them could provide the benefits SCO touts of putting Unix on PC hardware. A decade ago, there was already a fairly diverse PC Unix market.

But by the mid 90s market consolidation in the PC Unix world was already well underway. The Usenet discussion following the February 1995 "going out of business" announcement of one such PC Unix clone vendor, the Mark Williams Company behind the product “Coherent”, is a good sampling of what the PC Unix community understood eight years ago. The users discussing the issue knew exactly why Coherent went under. Already, Linux was a worthy substitute for traditional PC Unix systems, and the Unix clone market was seen as shrinking due to it. The rationale behind open-source development was already fairly mature, as expressed in this post. There was plenty of warning that the proprietary Unix market was in the process of being commoditized by open-source software. SCO had every chance to see its fate coming, and SCO/Tarantella eventually did, which is why it sold its operating systems division to SCO/Caldera.

The history of AT&T's archaic Unix has, for over a decade now, been an example of the “bigger fool” theory. Buy it, then find a bigger fool to buy it from you. AT&T got rid of it because they could not make money from it. Novell chose to exit the Unix market in 1995. SCO chose to exit the Unix market in 2000. Caldera's current lawsuit can be seen as its own form of exit strategy: they want to be purchased, and are making as much noise as they can until somebody buys them to shut them up.

40. The simplicity and power of this “UNIX on Intel” business model helped SCO grow rapidly. SCO gained other large enterprise customers such as CitiGroup, K-Mart, Cendant, Target Stores, Texas Instruments, Walgreens, Merck, Sherwin Williams, Radio Shack, Auto Zone, British Petroleum, Papa John's Pizza, Costco and many others.

When PC hardware became cheap and ubiquitous, SCO/Tarantella failed to capture even 1% market share on it; over 90% of all PC systems shipped with DOS or Windows, leaving Xenix lumped in with also-rans like OS/2 and CP/M-86.

Once again, SCO tries to confuse the meaning of "enterprise computing". SCO has sold cheap low-end systems to large enterprises, but the use they're putting SCO's products to does not qualify as "enterprise computing". Selling napkins to a fancy restaurant does not make one an expert on French cuisine.

41. As Intel's prominence grew in the enterprise computing market, SCO's early version of OpenServer also grew into the operating system of choice for enterprise customers who wanted an Intel-based computing solution for a high volume of repetitive, simple computing transactions.

With no timeframe here, it's hard to tell what they're talking about. With the introduction of the 32-bit 80386 (I.E. 386) processor in late 1985, the x86 line stopped being purely a 16-bit toy, but until IBM first used the 386 in a PC the existence of the processor had little impact on the PC market.[7] IBM announced the availablility of its first 386 processor in the PS/2 line, in April 1987, at the same time it announced AIX for the PS/2, which was a 386 product. SCO shipped Xenix 386 in 1987, and Sun also produced a 386 version of its BSD based SunOS, (before Solaris X86).

42. SCO OpenServer is based on the original UNIX Software Code developed by AT&T, but was modified by SCO for the functionality described above. Thus, while performing single-function applications, SCO OpenServer did so, and continues to do so, with the 99.999% reliability of UNIX.

Translation: Xenix (which became OpenServer) was based on an early version of AT&T's Unix, pared down for early 16-bit IBM PC systems at the behest of Microsoft. It worked about as well as DOS did on the same hardware for running "single-function applications".

43. Over 4,000 separate applications have been written by developers around the world specifically for SCO OpenServer. Most of these applications are vertical applications for targeted functions, such as point-of-sale control for specific industries, inventory control for specific industries, and related functions.

These numbers are unimpressive. Even IBM's OS/2 (a niche operating system IBM gave up on years ago) had more than 4,000 separate applications; DOS and Windows are an order of magnitude beyond that. Linux currently has twice that many open-source projects listed as "Production/Stable" or above in the Trove software map alone.

More to the point, consider SCO/Tarantella's own behavior on the application front before Caldera even bought them. The 86Open Project was an effort to define a common binary executable format to allow the same programs to run without modification on all the various x86 Unix variants. The first meeting of the group took place at SCO's corporate headquarters, on August 22, 1997. Shortly thereafter, a SCO developer named Michael Davidson wrote the utility lxrun, which allowed Linux utilities to run on OpenServer and UnixWare. Davidson made lxrun an open-source project, which allowed Sun to port lxrun to Solaris, and on July 25, 1999 the 86Open group declared victory and dissolved, with the Linux executable format now being the lingua franca of the x86 Unix world.

SCO/Caldera continues to offer lxrun to its customers, along with many hundreds of megabytes of other open-source code, on the SCO Skunkware page. Note that SCO Skunkware was a project of SCO/Tarantella, and predates the Monterey project by a number of years.

As far as SCO/Caldera is concerned, lxrun has been obsoleted by a new product, Linux Kernel Personality for UnixWare, a more comprehensive way to run Linux applications under UnixWare (and presumably a violation of the Linux trademark if they didn't include any Linux code, and a violation of the GPL if they did). It comes built into Open Unix 8.

The Linux Kernel Personality FAQ states:

The LKP feature in Open UNIX 8 represents a significant milestone in SCO's commitment to unify UNIX with Linux for business. Now, the huge number of Linux applications can be deployed on Open UNIX 8. Integrators and users do not have to compromise business requirements because the application they need is not supported on the platform of their choice.

The work to merge UnixWare and Linux predates Caldera's acquisition of SCO; the original Santa Cruz Operation was responsible for the headline SCO revamps UnixWare with Linux features in February 1999. That article contains some interesting quotes about Linux from SCO executives:

"So far as we've seen it's actually helped us," said Greg Schwarzer, director of small and medium business marketing at SCO. "Linux has got the word out that Unix on Intel is a viable alternative to Microsoft."...

Linux, meanwhile, has reinvigorated the Unix market, he said. "It has given a fresh, revitalized look to what people could do in the Intel space. Linux is a Unix movement. Revenues are going up strongly," Schwarzer said.

Adding support for Linux was a pragmatic choice. "We wanted our users to be able to take advantage of a lot of those applications being written for Linux," he said.

The SCO OpenServer Libraries

Paragraphs 44 through 47 talk about shared libraries, but do not seem to allege that these libraries are either distributed with Linux systems or commonly pirated; nor is there any claim that IBM has misappropriated them. Their main complaint seems to be that nobody wants to buy their SCO emulation module, ergo it must be widely pirated.

44. Much of the functionality of an operating system is made available to application developers by means of “libraries” of code that are supplied by the operating system vendor. These libraries contain many “functions” or “routines” which can be used by application developers to perform various common tasks such as reading or writing a file or opening a new window on the screen.

45. SCO OpenServer, as with many other operating systems such as Microsoft Windows, makes use of a special kind of library called a “shared library.” The code for all of the routines in a particular shared library is stored in a separate file, and this code is loaded into memory “on demand” when an application needs to make use of it. There are several benefits that come from using “shared libraries” - applications can be smaller and use less memory because a single copy of the library code is “shared” by all of the applications that make use of it, and system vendors can easily update the library code in order to fix problems or provide enhanced functionality. A side effect of this is that it is also very easy to make a copy of a shared library.

46. In creating the thousands of SCO OpenServer Applications, each designed for a specialized function in a vertical industry, software developers wrote software code which specifically made use of the SCO OpenServer shared libraries (hereinafter the “SCO OpenServer Shared Libraries”), and thus the presence of the SCO OpenServer Shared Libraries on a particular system is required in order for these applications to be able to run and function correctly.

Without application programs that give an operating system something to do, there's not much point to having most computers even switched on. Third parties wrote a lot of valuable applications, and years ago before Linux rose to prominence and back when the status of BSD was still legally uncertain, they deployed a lot of them on systems like Xenix (which became SCO OpenServer).

When a third party's source code was compiled to create binary executable files to run on a SCO platform, those binaries would try to load SCO's shared libraries to access standard operating system functions. When compiled on another platform (like BSD or Linux), those binaries would instead expect to link up with the BSD or Linux shared libraries. (Versions could even be compiled for other operating systems that had a Unix compatability layer, such as the EMX package for OS/2 or the Cygwin package for Windows. In this case, the resulting binary file would depend on the shared libraries of the appropriate Unix emulation package.) This is an incidental dependency, not inherent in the third party software but simply an artifact of creating binary executable files intended to run on a specific platform.

47. Linux offers a “SCO emulation module,” originally called “iBCS” and now known as “linux-abi” which enables applications which were originally developed to run on SCO OpenServer to be run on Linux. However, in order for these applications to function, the SCO OpenServer Shared Libraries must also be copied onto the Linux system. The SCO OpenServer Shared Libraries are the proprietary and confidential property of SCO. SCO OpenServer has been licensed to numerous customers subject to restrictions on use that prohibit unauthorized use of any of its software code, including without limitation, the SCO OpenServer Shared Libraries. SCO does not give permission for copying of the Shared Libraries for use outside OpenServer without payment of separate licensing fees.

IBCS stands for the "Intel Binary Compatibility Standard". It was yet another attempt (like 86open and other efforts) to unify the fragmented and incompatible Unix-on-x86 market. Intel pushed the standard in the early 1990s and it's been dead for so long that most references to it have fallen off the web. You can still find a FAQ about it, and an article from 1999 described it as follows:

During the late 1980s/early 1990s Intel promoted a standard binary format for UNIX executables, known as iBCS2 (Intel Binary Compatubility Standard 2). Linux has a kernel module available which provides support for all the various derivatives of iBCS2, so that binaries compiled for a whole range of older UNIXes can run: these include i386 BSD (386BSD, FreeBSD, NetBSD, BSDI/386), SVR4 (Interactive, Unixware, USL, Dell.), generic SVR3, SCO 3.2.x (COFF), SCO OpenServer 5 COFF and ELF binaries, Wyse V/386 (SVR3 with extensions), Xenix V/386, and Xenix V/286. This means that the odds are if you have an existing PC-based UNIX system running some important software (say, an arcane stock management system that isn't supported any more) you may well be able to get it to run under Linux.

Because SCO/Tarantella supported iBCS in UnixWare, it was sometimes possible to run UnixWare applications under other Intel Unix versions, such as Linux, using an iBCS support package for the platform. This would be the only reason to install SCO shared libraries on a Linux platform ala paragraph 46, and in some cases that still wasn't necessary. Linux did have its own shared libraries, and since SCO was presumably providing a standard POSIX/SUS Unix API, in theory adapting them shouldn't be too difficult. In practice, the debugging work required generally wasn't worth the effort, since these deployments were normally aids to migration (to be used only until a native Linux version of the application became available) rather than a permanent solution.

The reason iBCS for Linux has fallen out of use is that it was mainly used to run a small number of proprietary applications that have since released native Linux versions, such as Netscape (which elevated Linux to Tier 1 platform status around the time it, announced its intention to release source code in 1998), and SAS (which ported its software to Linux in 2000). The main iBCS application that Linux users wanted to run, by a wide margin, was Oracle's database product, and when Oracle announced support for Linux in July 1998, iBCS rapidly became irrelevant. In the past 5 years, far more new proprietary packages have been released for Linux than for both of SCO's proprietary Unix platforms combined.

Oracle's announcement of Linux support came the same month Linux creator Linus Torvalds was on the cover of Forbes. As the leading database vendor, Oracle's announcement prompted IBM to announce a Linux version of its own database, DB2, two months later. That announcement was IBM's first official acknowledgement of or support for Linux at the corporate level[8], and it was a reactive rather than proactive measure.

In 2003, SCO no longer has interesting applications that Linux lacks. Since at least 1998, before IBM first became even peripherally involved with Linux, major software development houses have shipped Linux versions of their software, often instead of SCO versions. SCO has responded by creating technologies like lxrun, and the Linux Kernel Personality for UnixWare, to run Linux applications on SCO's operating system. SCO/Tarantella realised this was a losing strategy and got out of the proprietary Unix business in 2000.

Recently, SCO resurrected the issue of iBCS support as part of its "System V for Linux" product, the first deliverable of the SCOsource initiative, which is the newly launched division of SCO that is driving the lawsuit against IBM. The presentation describes System V for Linux as "SCO's Shared UNIX Libraries from OpenServer and UnixWare for use with Linux", and claims that two answers to the question "Why License SCO's Intellectual Property" are "Strengthen Linux by licensing value-add IP" and "Increase UNIX application use on Linux". It also claims:

In the late 1980s, SCO created an open specification called ibcs2 (Intel Binary Compatibility Standard) to allow UNIX applications to run on Intel standard hardware using SCO's shared libraries.

Because ibcs2 is an open specification, the Linux community was able to freely copy this and rename it Linux ABI or Linux Application Binary Interface.

In order to run UNIX apps on Linux, customers must have Linux ABI and SCO's shared libraries.

SCO's shared libraries could never be licensed outside of the SCO operating systems until today.

The third paragraph of that is wrong: in order to run UnixWare applications under Linux, SCO's shared libraries might be necessary. When attempting to run applications for other x86 versions of Unix, such as the x86 versions of AIX or Solaris, SCO's shared libraries would be of no use whatsoever.

It's interesting to note that they claim to have invented the Intel Binary Compatibility Standard, rather than Intel. Also, the most recent document they can find explaining how to run a SCO binary on Linux explains how to run WordPerfect 5, from 1995. Caldera paid to have WordPerfect 6.0 ported to Linux, which is why subsequent versions have also been available for Linux. Caldera's presentation also references the retention of Boies, Schiller and Flexner by SCOsource (the lawfirm retained to persue the current suit against IBM) while at the same time it hawks a Linux product. Apparently, this attempt to profit from Linux and the IBM lawsuit were launched more or less simultaneously.

SCO points out that System V for Linux is the first time they've licensed their shared libraries seperately from their operating system, but skips over the fact that people who legally owned a copy of a SCO operating system but chose to migrate to Linux could legally use those shared libraries under Linux, because they do have a license to own them and copyright law cannot pevent the re-use under the “first sale” doctrine. (It could also be considered an example of "space shifting".) The only users likely to have a significant number of legacy SCO applications they'd want to run on a new platform are users who migrated from a SCO platform to Linux, and thus have a SCO license they are otherwise no longer using.

SCO's Development of UnixWare on Intel

48. While the original SCO OpenServer operating system performs with all the reliability and dependability of other UNIX systems, it was originally designed for the initially low processing power of Intel chips. Therefore, SCO OpenServer does not offer the same level of multiprocessor capabilities that other versions of UNIX offer.

So why did they spend so many paragraphs insisting it's an "enterprise operating system"?

49. During or about 1992, SCO's predecessor in interest, Novell, Inc. (“Novell”), acquired from AT&T all right, title and interest in and to the UNIX software code, the AT&T Software and Sublicensing Agreements, the copyrights and related and ancillary products for $750 million in Novell stock. For branding purposes, Novell renamed UNIX as “UnixWare.”

Novell acquired Unix System Labs (USL), which AT&T had spun off as an independent company. Novell bought out AT&T's stake in USL. Novell didn't acquire "all right, title and interest" to Unix from AT&T because AT&T didn't have it to give: a large number of existing licenses had been granted to companies like IBM delegating rights to create derived works that belonged to the licensees, not to AT&T. For years, far more Unix development had taken place outside of AT&T or USL than within it.

This period in Unix history is still remembered for fragmentation, divergence, and increasing incompatibility, because licensees were not sharing new developments with each other. One reaction against this was standardization efforts like Posix, OSF/1, the X/Open portability guides, the Single Unix Specification, and The System V Interface Definition. Most modern unix-like systems, including Linux, are compliant with many or all of these published standards. Technically speaking, even Microsoft's Windows NT has achieved nominal compliance with some of them.

The largest development branch AT&T didn't own was from the University of California at Berkeley. The BSD lawsuit was settled on Novell's watch, and settled in Berkeley's favor, not only because Berkeley exercised their right to create derived works until they had replaced the last of the original System V code, but also because AT&T and USL had contaminated System V with code they didn't have rights to. Derived code belonging to licensees wound up in System V, and it came out in court. Novell ordered USL to settle the lawsuit, and unloaded System V shortly afterwards.

50. On or about September 19, 1995 the Santa Cruz Operation acquired all right, title and interest in and to UNIX and UnixWare source code, the AT&T Software and Sublicensing Agreements, the copyrights, claims arising after the closing date against any party and all related and ancillary products and rights from Novell, excepting only the right to certain existing ongoing royalty payments which was retained by Novell.

SCO did not get the Unix trademark, which Novell gave to The Open Group, and Novell has disputed the copyright assignment (a question which simply had not arisen for seven years between the transaction and this lawsuit). SCO also continued to pay a significant amount of royalties to Novell, $19 million in the first quarter alone. Novell's 1996 first quarter 10-Q records the transaction as follows:

In December 1995, Novell sold its UNIX and UnixWare product line to the Santa Cruz Operation, Inc. (SCO). The Company realized a small gain and recorded $19 million of royalty revenue from this transaction in the first quarter of fiscal 1996. Under the agreement, Novell received approximately 6 million shares of SCO common stock, resulting in an ownership position of approximately 17% of the outstanding SCO common stock. The agreement also calls for Novell to receive a revenue stream from SCO based on revenue performance of the purchased UnixWare product line. This revenue stream is not to exceed $84 million net present value, and will end by the year 2002. In addition, Novell will continue to receive revenue from existing licenses for older versions of UNIX System source code.

The "older versions", on which SCO still pays royalties to Novell, include existing System V licensees. SCO/Caldera's most recent annual report reports a royalty payable to Novell of $1.4 million for fiscal year 2002. If SCO has free and clear rights to Unix, why are they still paying royalties on it to Novell?

51. From and after September 1995, SCO dedicated significant amounts of funding and a large number of UNIX software engineers, many of whom were original AT&T UNIX software engineers, to upgrade UnixWare for high-performance computing on Intel processors.

52. By approximately 1998, SCO had completed the majority of this task. That is to say, UnixWare had largely been modified, tested and “enterprise hardened” to use Intel-based processors in direct competition against IBM and Power PC chips, the Sun SPARC chip and all other high-performance computing UNIX platforms for all complex computing demands. The term “enterprise hardened” means to assure that a software product is fully capable of performing under the rigorous demands of enterprise use.

Translation: SCO/Tarantella invested three years of work to try to breathe life into the obsolete product Novell had just unloaded on them. They tried to upgrade it to some sort of relevance in the marketplace, and the result was to turn UnixWare 2 into UnixWare 7, which the industry perceived as a non-event. In 1999 The UnixWare "Business Edition" was released, and a quote from an article about it states, "The Business Edition supports a single-CPU system with up 4 gigabytes of memory and up to five users." During this period, top of the line x86 "Xeon" systems from Intel already supported up to 8 processors and 64 gigabytes of memory.

In 1998, just as SCO/Tarantella stopped throwing good money after bad, Linus Torvalds appeared on the cover of Forbes magazine, Oracle announced support for Linux, Netscape released its source code (citing Linux as an inspiration), and Linux-based "Beowulf" clusters showed up in the top 500 supercomputers list and were covered by CNN. All of this occurred before September, when IBM first supported Linux — not by contributing directly to development of the operating system, but by releasing a Linux version of IBM's DB2 database application in response to database market leader Oracle's Linux initiative (which has continued). Linux hit the big time long before IBM noticed it.

SCO/Tarantella had noticed Linux as early as 1995, when Linux put Coherent out of business. In 1998 its responses to Linux included iBCS, lxrun, and the 86open project. It also stepped up its "skunkware" project, to collect and distribute open source for SCO platforms (even shipping skunkware CD's with SCO's operating system). SCO's "Open Source Program Architect", Ronald Joe Record, gave a very interesting slide presentation about SCO's open-source efforts at SCO Forum 98. Major releases of SCO Skunkware were made in 1998, 1996, and earlier.

SCO/Tarantella's co-founding of the Monterey project in October 1998 was a second attempt to make their proprietary Unix technology relevant to enterprise customers. If their own in-house efforts had succeeded, they would not have needed to partner with IBM, Sequent, and others.

The three years of development work in pargraphs 51-52 have no bearing on the historical System V code other vendors had already licensed and upgraded on their own in the 1980s and early 90s. Any development work done by SCO in 1995 would merely create yet another derivative work, in this case UnixWare 7.

53. SCO was ready to offer large enterprise customers a high-end UNIX computing platform based on inexpensive Intel processors. Given the rapid growth of Intel's performance capabilities and Intel's popularity in the marketplace, SCO found itself in a highly desirable market position. In addition, SCO still had its SCO OpenServer business for retail and inventory-targeted functions, with its 4,000 applications in support.

54. Prior to the events complained of in this action, SCO was the undisputed global leader in the design and distribution of UNIX-based operating systems on Intel-based processing platforms.

This statement is simply false. See the original OSI position paper for a detailed refutation. By the end of 1998 SCO had added support to its products for the Intel Binary Compatibility Standard (iCBS), participated in the 86open project, and created lxrun. None of these actions would have made sense for an “undisputed leader”.

Project Monterey

55. As SCO was poised and ready to expand its market and market share for UnixWare targeted to high-performance enterprise customers, IBM approached SCO to jointly develop a 64-bit UNIX-based operating system for a new 64-bit Intel platform. This joint development effort was widely known as Project Monterey.

The translation of "poised and ready to expand" is something like "scratching its head wondering why nothing had happened". SCO/Tarantella had never previously been in the same league as Sun, HP, and IBM, and its development efforts did not suddenly elevate it there. Linux was already eating into proprietary Unix vendors' marketshare on 32-bit x86 machines, prompting Sun to give away free licenses to Solaris for x86. And Linux itself was not the only reason for the erosion of SCO's market; the success of Linux and FreeBSD helped pave the way for Apple's may 1998 announcement of a new OS development strategy, MacOS X based on a combination of FreeBSD and Carnegie Mellon's Mach microkernel. (Here's an article on the convoluted history of MacOS X). Apple's move to Unix also cut into the low and midrange Unix markets, SCO's bread and butter. Their attempted move to high end "enterprise computing" was an exit strategy from their rapidly deteriorating existing market.

SCO/Caldera seems to be attempting to imply that project Monterey was some kind of bait distracting them at a crucial moment, and that if it hadn't been for IBM, UnixWare would have blossomed into a success. The fact that UnixWare continued to be developed and marketed during Monterey development and continued to be a dud is a counter-argument. SCO/Caldera joined Monterey because their own in-house development efforts had not achieved significant enterprise market share. They wanted IBM's help. They jumped at the chance to get it.

56. Prior to this time, IBM had not developed any expertise to run UNIX on an Intel processor and instead was confined to its Power PC processor.

Paragraph 56 is completely wrong. AIX for PS/2 was announced at the PS/2 launch in 1987. IBM also shipped AOS, a BSD derivative, for PS/2 systems. IBM's first Unix product for its x86 PC shipped in 1983 (as described in the response to Paragraph 26).

In addition, the statement is intentionally misleading. Although Itanium was from Intel, it was a very different design from Intel's traditional x86 processor line. Monterey was an effort to port Unix to a completely new hardware architecture, a 64-bit processor with a new instruction set (based on VLIW design principles which, like RISC, IBM had invented years before). Itanium (originally code-named Merced) was jointly designed by Intel and HP. Writing code for the 64-bit Itanium processor has more similarity with writing code for 64-bit RISC Power PC processors than writing for 32-bit CISC x86 processors.

Monterey was not the first such 64-bit processor either. The first 64-bit RISC processor, the DEC Alpha (now owned by HP due to its purchase of Compaq, which had previously purchased DEC) was introduced in February 1992. The first 64 bit Unix for this processor, DEC OSF/1 (later renamed Tru64 Unix) was introduced in March 1993. It had nothing to do with SCO.

The article 27 years of IBM RISC points out that IBM's own internal development led to the release of a 64-bit PowerPC processor and a 64-bit version of AIX (4.3), in 1997. The article says:

The RS64 (also known as Apache) is the first 64-bit PowerPC RISC processor (October 1997). The RS64 is a superscalar processor optimized for commercial workloads... IBM brings 64-bit technology to the market introducing the RS/6000 Enterprise Server model 7017-S70 (125 MHz, code named Raven), the first 12-way SMP system, and AIX Version 4.3.

IBM had more experience with 64-bit computing than SCO did, having designed and marketed its own 64-bit processor and a version of AIX that could run 64-bit applications.

57. SCO, on the other hand, had over 15 years of expertise in adapting UNIX to Intel based systems. Moreover, SCO had spent the previous 18 months working closely with Intel to adapt its existing UnixWare product to work on the new 64-bit Intel processor. That project, known as "Gemini-64," was well underway when work on Project Monterey was started. In furtherance of, and in reliance on, IBM's commitment to Project Monterey, SCO ceased work on the Gemini-64 Project and expended substantial amounts of money and dedicated a significant portion of SCO's development team to Project Monterey. Specifically, plaintiff and plaintiff's predecessor provided IBM engineers with valuable information and trade secrets with respect to architecture, schematics, and design of UnixWare and the UNIX source code for both 32- and 64-bit Intel-based processors.

Over 15 years ago SCO was adapting Xenix to the 16-bit 8086 and 286 processors in the original IBM PC and AT lines, which were less powerful than a modern cell phone. This has no relevance whatsoever to modern "enterprise" systems. The first 32 bit processor in the x86 line, the 386, supported over a dozen Unix ports, including AIX for the PS/2 from IBM.

SCO is trying to confuse "Gemini 64" with the three years of development they mentioned in pargraphs 51-52. The purpose of the development in pargraphs 51-52 was to turn UnixWare 2, which SCO/Tarantella purchased from Novell in 1995, into UnixWare 7, released in 1998. This development resulted in a product release for 32-bit x86 Intel processors (the 386, 486, and Pentium lines).

The Gemini-64 project was an experimental effort at SCO to play around with Intel's unfinished 64-bit design (code named "Merced", and released as "Itanium"), which could only be tested on emulators since Intel hadn't manufactured any actual 64-bit processor chips yet. SCO eventually admitted they were in over their heads, canned Gemini-64, and sent the Gemini engineers off to work with IBM.

By the way, does SCO really claim to have trade secrets related to the architecture and schematics of Intel's new hardware design? (Wouldn't these be Intel's trade secrets, which IBM could get from Intel? If IBM could NOT get them directly from Intel, wouldn't SCO be violating its agreements with Intel in passing them on?) The most charitable reading is that SCO is trying to inflate its importance, or simply confusing two issues again.

58. By about May 2001, all technical aspects of Project Monterey had been substantially completed. The only remaining tasks of Project Monterey involved marketing and branding tasks to be performed substantially by IBM.

Except that market interest in Monterey never materialized, in large part because the Itanium hardware was a flop, completely unsuccessful in the mainstream computer market. To read the gory details of the Itanium disaster, see Appendix D at the end of this document.

Beyond that, the initial point of the Monterey project was another standardization effort, like the Open Software Foundation of the late 1980s. Periodically, when the proprietary Unix market gets too fragmented, a group of vendors band together for a while, and after a while collapse into bickering and pointless infighting. The Open Software Foundation was originally a joint effort between leading Unix vendors including IBM, Hewlett Packard, and Digital Equipment Corporation (now part of HP, via Compaq), in reaction to AT&T's purchase of an equity stake in Unix vendor Sun Microsystems. (Everybody but Sun and AT&T formed an opposing camp.) OSF eventually collapsed, and their operating systems release (OSF/1) became Tru64 Unix, a product for the DEC Alpha processor. Other theoretically similar standardization efforts predating Monterey include Unix International, and the groups behind Posix and the Single Unix Specification. (The second edition of the book "Life with Unix" talks about all this in detail.)

Monterey was another such standardization effort, as explained in Appendix D.

59. On or about May 2001, IBM notified plaintiff that it refused to proceed with Project Monterey, and that IBM considered Project Monterey to be “dead.”

For entirely justifiable business reasons. The Itanium was a disaster, IBM had purchased Sequent for its NUMA technology, and Linux had gone mainstream around 1998. Project Monterey was an obvious failure long before May 2001.

Three years ago, the lead author of this commentary, Rob Landley, summarized the Monterey situation in a comment attached to a July 19, 2000 article, which was one of many articles pointing out that the reason the Linux distributor Caldera purchased SCO was not to get its technology, but to get an international distribution channel through which to sell Linux. (The fact that Caldera immediately renamed itself "Caldera International" might have been another hint about what part of the deal excited them.)

The AT&T UNIX Agreements

60. AT&T Technologies originally licensed the UNIX operating system software code to hundreds of software licensees, including defendant IBM, for the UNIX operating system software source code, object code and related schematics, documentation and derivative works (collectively, the “UNIX Source Code”). To protect the confidential and proprietary source code information, these license agreements, as detailed below, contained strict limitations on use and distribution of UNIX source and binary code.

SCO/Tarantella contributed code to Linux in 1999, and was thinking of putting out its own Linux distribution until Caldera purchased its operating systems division. Some news reports state that SCO/Tarantella's moves into Linux were one reason for the Caldera acquisition.

SCO/Caldera was a Linux distributor, and repeatedly announced plans to put Unix technology into Linux. What if one of Caldera's Linux engineers believed them? Caldera employed a number of high-profile Linux engineers, such as Christoph Hellwig, who was in the top ten list of contributors to Linux 2.4 while working for Caldera in Germany.

Christoph Hellwig is of the opinion that SCO probably didn't intentionally put UnixWare code into the Linux kernel:

"Especially given that the kernel internals are so different that you'd need a big glue layer to actually make it work and you can guess how that would be ripped apart in a usual lkml [Linux Kernel Mailing List] review :)"

I.E. he believes that SCO/Caldera had "better things to do than trying to retrofit UnixWare code into the linux kenrel[sic]" because if SCO/Caldera had submitted UnixWare code to the Linux kernel, it would have been rejected on technical grounds. (This is further support for the contention that UnixWare is obsolete crap.)

Christoph remains very active in Linux kernel development since having left SCO/Caldera, and is only one of thousands of people outside of IBM who have legally worked with traditional proprietary Unix code, but have left proprietary Unix behind in favor of Linux. SCO claims that code licensed to thousands of entities by many different companies over a period of decades has never legitimately leaked once. The co-author of this document, Eric Raymond, doesn't think so: his No Secrets page has already collected over 100 reports from old-time Unix administrators who still have copies of System V-derived Unix source code in their possession.

Hundreds of thousands of developers have had legitimate access to this code for two decades, and SCO claims that no employment agreement ever missed telling them to destroy their copies when they left the company, no company ever went bankrupt owning Unix code and had its unformatted hard drives liquidated at auction, no disks or printouts were ever thrown out without being shredded and fished out of the trash...

AT&T sold educational licenses for Unix source code to universities at a considerable discount over commercial use, yet how many professors made their students sign a non-disclosure agreement? How can something explicitly licensed for educational purposes and knowingly taught to all comers in a university classroom be considered a "secret", by any definition?

And with all these thousands of different organizations that had access to this code, why is SCO picking on IBM specifically? (Other than IBM's deep pockets?)

61. When SCO acquired the UNIX assets from Novell in 1995, it acquired all right, title and interest in and to the UNIX operating system technology, including all claims against any parties relating to any right, property or asset used in the business of developing UNIX and UnixWare. As a result of this acquisition, SCO became the authorized successor in interest to the original position of AT&T Technologies with respect to all licensed UNIX software products.

Just like the DR-DOS lawsuit in 2000. None of Caldera's predecessors felt like suing, but Caldera buys up dead technologies to see if it can squeeze a lawsuit out of them. Microsoft was motivated to settle by the DOJ antitrust loss. IBM does not have a history of misappropriating code. (And Unix is one of thousands of different technologies they've licensed from third parties; the vast majority of these third parties seem happy.)

62. There are two primary types of software licensing agreements between AT&T Technologies and its various licensees:

a) The AT&T-related software agreements are collectively referred to hereinafter as the “AT&T UNIX Software Agreements.”

b) The AT&T-related sublicensing agreements are collectively referred to hereinafter as the “AT&T UNIX Sublicensing Agreements.”

The AT&T UNIX Software Agreements and the AT&T UNIX Sublicensing Agreements are sometimes collectively referred to hereinafter as the “AT&T UNIX Agreements.”

But the interesting one here is the agreement with IBM. We covered SCO Exhibit C in the rebuttal to paragraphs 2 and 6 of the complaint, which show that IBM has ownership of its own derivative works and that it can use what it learns elsewhere.

SCO Exhibit D contains "Amendment X", which states:

Upon payment to SCO of the consideration in the section entitled "Consideration", IBM will have the irrevocable, Fully paid-up, perpetual right to exercise all of its rights under the Related Agreements beginning January 1, 1996 at no additional royalty fee.

These rights include the rights recognized in Exhibit C.

63. Plaintiff is successor in interest to, and owner of, all contractual rights arising from and related to the AT&T UNIX Agreements.

The contracts stipulate that they pay ongoing royalties to Novell. Novell retained the Unix patents, and gave the Unix trademark to somebody else. The Unix intellectual property portfolio suffers from the same fragmentation and degradation the proprietary Unix market suffers from.

The IBM Related Agreements

64. On February 1, 1985, AT&T and IBM entered into certain AT&T UNIX Agreements:

a) Software Agreement Number Soft-00015 (“AT&T / IBM Software Agreement” attached hereto and incorporated herein as Exhibit A);

b) Sublicensing Agreement Number Sub-00015A (“AT&T / IBM Sublicensing Agreement” attached hereto and incorporated herein as Exhibit B).

65. AT&T and IBM have entered into a side letter on that date (“AT&T / IBM Side Letter” attached hereto and incorporated herein as Exhibit C).

66. In addition, AT&T and IBM have entered into nearly 400 supplemental agreements over the years, including Supplement No. 170 (Supplement No. 170 is attached hereto and incorporated herein as Exhibit D). Supplement No. 170 is the document that specifies the royalty amounts and computer CPUs upon which royalty amounts were due to be paid by IBM.

That is, IBM bought an irrevocable perpetual license like the kind SCO is trying to flog through SCOsource. They think somebody may have contributed old System V code to Linux, they think it might still be a defendable trade secret, they think it was IBM that did it, and they think that the whole world has been maliciously using their code even though they themselves shipped it without noticing for years. They also don't think it was their own engineers even though Caldera announced plans to merge their Unix and Linux product lines, and paid a full-time development staff to put code into the Linux kernel, and supported many other open-source projects.

It's even possible that SCO/Tarantella put some of their code into Linux, since they toyed with a Linux services business in 1999, and were at one point considering the release of their own Linux distribution.

67. Thereafter, Amendment X to Software Agreement SOFT-00015, as amended, was executed on or about October 16, 1996 by and among IBM, The Santa Cruz Operation, Inc. (“SCO”) and Novell, Inc. (“IBM Amendment X” attached hereto and incorporated herein as Exhibit E). Among other things, Amendment X effectuated a royalty buy-out by IBM pursuant to the royalty terms and amounts specified in Supplement No. 170.

SCO agreed to the word "Irrevocable". At the top of page 2 of Exhibit E they reserve the right to "enjoin or otherwise prohibit IBM from violating... SCO's rights", but if it's irrevocable they can't revoke the license. A more apt penalty would be an injunction asking for the removal of the offending code from the product in question, along the lines of the BSD settlement, without monetary damages.

68. Collectively these agreements, side letter and amendment are referred to hereinafter as the “IBM Related Agreements.”

The Sequent Agreements

69. On January 28, 1986, AT&T and Sequent (now an operating division of IBM) entered into certain AT&T UNIX Agreements:

a) Software Agreement Number SOFT-000321 (“Sequent Software Agreement” attached hereto and incorporated herein as Exhibit F);

b) Sublicensing Agreement Number SUB-000321A (“Sequent Sublicensing Agreement” attached hereto and incorporated herein as Exhibit G).

70. The Sequent Software Agreement and the Sequent Sublicensing Agreement are sometimes collectively referred to hereinafter as the “Sequent Agreements.”

IBM bought Sequent for its NUMA technology, which is hardware. The only software expertise of interest is the ability to program that hardware.

Sequent did not license the capability to create NUMA hardware from Caldera, Tarantella, Novell, or AT&T. It was original to Sequent. Any NUMA technology in Sequent's Unix is something Sequent put there.

Marketplace Value of UNIX

71. UNIX's value in the enterprise marketplace is largely a function of its reliability, extensibility, and robust performance capability. That is to say, it virtually never needs repair, it performs well under a wide variety of adverse circumstances, and it can be extended throughout an enterprise and across multiple processors to perform unified or disparate tasks in a seamless computing environment. Because of these features, UNIX-based equipment has replaced mainframe computers for all but the most demanding computing tasks. And, because UNIX-based equipment is far cheaper than mainframe computing equipment, a customer who cannot otherwise justify the cost of mainframe computers can otherwise gain the advantages of “supercomputing” operations through use of UNIX-based equipment.

Linux systems have been replacing other Unix systems since before the Mark Williams Company went out of business in 1995.

A representative story is how, in 1996, Cisco replaced its Solaris-based print servers with Linux systems. Originally the printing system, which handled invoices, customer orders, manufacturing documents, and other tasks without which the company could not even temporarily function, was considered too important to trust to Linux. But Cisco was growing faster than Solaris servers could be installed (factoring in budget approvals, purchase orders, and shipping). A few Linux servers were installed to handle the overflow, since a Linux system could be assembled and installed out of spare PC parts and downloaded software in a few hours. But after a 1997 power failure, the Solaris servers did not come back online (requiring manual maintenance to repair after being improperly shut down), and the Linux servers all came back online as soon as the power was restored, and handled Cisco's entire printing needs while the Solaris machines were out of commission. After Linux proved itself in this crisis, Cisco switched entirely over to Linux for its mission-critical print system and retired the Sun print servers on August 1, 1997.

The system that proved it "virtually never needs repair" in this instance was not proprietary Unix, but Linux.

As for Unix being "Extended across multiple processors", if SCO/Caldera means SMP support, it should remember that Caldera proudly sponsored the initial SMP work for Linux which went into the 2.0 kernel. (The 2.0 version of the Linux kernel was released June, 1996.) SCO still has a web page bragging about their contributions to the Linux kernel.

If, on the other hand, extending across multiple processors means that the same operating system not only runs on x86 processors, but is also available for Sun's Sparc, IBM's Power PC and S/390 mainframes, The Alpha and HP-PA processors owned by Hewlett Packard, and many others, then SCO's products fail where Linux succeeds. SCO's UnixWare and OpenServer are 32-bit x86 systems only. Back in 1999, Caldera was proud to announce a version of Caldera OpenLinux for Sun's 64-bit UltraSparc processors, but its Unix products (UnixWare and OpenServer) are not capable of running on that platform.

And if you're going to talk about supercomputing, remember that some of the 10 most powerful computers on the Top 500 supercomputers list are Linux-based Beowulf clusters. SCO machines aren't in the Top 500.

72. One or more of the different versions of UNIX-based operating systems sold by Sun, IBM, SCO, SGI, and others, is the operating system of choice for large enterprise computing operations in virtually 100% of the Fortune 1000 companies.

Take SCO out of that picture and the answer's the same. SCO is NOWHERE in "large enterprise computing operations".

73. UNIX gained this prominence in the computing marketplace because of twenty years of development and over one billion dollars invested by plaintiff and its predecessors to create a stable, reliable operating system to perform the mission critical work required by large enterprises.

The majority of this investment took place before Novell purchased USL, and has been depreciating on a schedule driven by Moore's Law for at least 15 years.

IBM pumped more than $1 billion into OS/2. Microsoft pumped more than $1 billion into DOS. The VMS operating systems for VAX minicomputers is another such example. All of these are obsolete relics.

The investment by companies like Sun, IBM, and HP dwarfs that of the plaintiff and predecessors. Sun's core business was selling Unix machines; this was not the case for AT&T or Novell. SCO/Tarantella was never a member of the Fortune 500, and neither was Caldera either before or after purchasing SCO. Sun, IBM, and HP are all Fortune 500 companies. The tail is trying to wag the dog here.

74. The recent rise of the global technology economy has been powered in large part by UNIX. Virtually every mission critical financial application in the world is powered by UNIX, including electronic transfers of funds. Real time stock trades are powered by UNIX. Inventory controls and distributions are powered by UNIX. All major power grids and all major telecommunications systems are powered by UNIX. Many satellite control and defense control systems are powered by UNIX. Virtually every large corporation in the world currently operates part or all of its information technology systems on a UNIX operating system.

1) None of that is SCO, it's all other Unix vendors. SCO was fighting for the scraps of the webserver business, and the majority of its sales come from cash registers.

2) Proprietary Unix had largely fragmented itself into irrelevance by the mid 1990s. SCO/Tarantella's Greg Schwarzer, director of small and medium business marketing, said in February 1999, "Linux has got the word out that Unix on Intel is a viable alternative to Microsoft." Proprietary Unix had been reduced to such irrelevance that SCO's own director of marketing needed outside help (from Linux) to convince people it was even a viable alternative!

Due to the infighting and fragmentation in the Unix market, the cover of the September 1992 issue of Byte magazine was devoted to the question, "Is Unix Dead?" A 2000 editorial in Software Development Times follows up on Byte magazine's question, reviewing some of the history before concluding that Linux was revitalizing the Unix world.

Caldera was one such Linux vendor until quite recently, when they chose to go down a proprietary cul-de-sac and revisit ancient mistakes.

Another example of how far proprietary Unix had fallen is that when IBM purchased Sequent, Sequent's NUMA systems were primarily running Windows NT, and most systems from Sequent's main competitor (Unisys) were running Windows 2000.

75. Based on its value in the marketplace, UNIX has become the most widely used and widely accepted operating system for enterprise, institutional and manufacturing applications throughout the world.

Yet SCO's Linux was not widely used, nor widely accepted. An article from ITWeek analysing Caldera's acquisition of SCO, quoted SCO/Tarantella's president on the reasons for the sale:

Tarantella's independence means two things. Firstly, it has to build a prestigious portfolio of big clients. "Our main objective is credibility," explained Mike Orr, Tarantella's president. "We are practically a startup now, so we need to get the big enterprise customers on board quickly."

Secondly, independence means Tarantella no longer needs to push SCO/Caldera products. The first choice among prospective enterprise customers is Sun Microsystems, after which they opt for Hewlett Packard, he admitted.

The predominant platform for the enterprise market, which Caldera expressly bought the two SCO divisions to sell into, is on Risc-based systems. "You'd be amazed how many sales of hardware are tied up with software," said Orr.

In the Linux community itself, Caldera was barely even perceived as a serious player.


76. Linux, or “GNU/Linux,” is an operating system variant or clone of UNIX System V Technology. According to leaders within the Linux community, Linux is not just a “clone,” but is intended to be a successor to UNIX System V. Linux, unlike UNIX, is distributed without a fee to its users.

Linux was created without reference to System V. Linux is indeed used to replace proprietary Unixes, but so was Coherent, and Linux put that company out of business in 1995.

77. As long as the Linux development process lacked central coordination, its direction was primarily aimed at meeting the computing needs of the Linux programmers themselves. As such, it posed little or no practical threat to SCO or to other UNIX vendors since the Linux developers did not have access to sophisticated high-end enterprise class multiprocessor systems, nor did they have any particular interest in supporting such systems.

Open source has not only proven itself in the real world, but has been mathematically determined to be a superior model of software development in the paper Closed source versus open source in a model of software bug dynamics by Damien Challet and Yann Le Du. The development process of Linux therefore made it a ‘practical threat’ from the beginning. SCO's assertion here is a disingenuous attempt to creartte a distinction between the threat level of pre-IBM and post-IBM Unix which is not supported by the actual facts.

As mentioned in the response to paragraph 71, Caldera itself sponsored some of the first Linux SMP work. In 1995 and 1996, the Linux community was already investigating the difference between commercial and scientific supercomputing. Later, when Linus Torvalds was hired to work at Transmeta in 1996, the first task they put him to was improving SMP support for Linux, which they used extensively. By mid 1997, Linus was working exclusively on SMP machines, as noted in this 1998 interview about the new 2.2 Linux kernel:

Marjorie: You seem to be quite excited about the addition of SMP--tell us about it. Will having this feature open new doors for Linux?

Linus: We had SMP support in 2.0 too, thanks mainly to Alan Cox. The new thing about 2.2 is that the SMP support is no longer just a tacked-on feature no one actually trusts, but is much better integrated. I've done all my development on SMP machines for the last year and a half.

Support on processors other than 32-bit x86 systems can be traced back to Jon "Maddog" Hall who used to be in charge of Unix development at Digital Equipment Corporation (later purchased by Compaq, and now a part of Hewlett Packard). In 1994, Jon Hall met Linus at the 1994 DECUS conference, and shortly afterwards convinced his company to donate an Alpha processor system to Linus Torvalds. This system sparked the first non-Intel port in Linus's tree[9], and the first port of Linux to a 64-bit system (hence making later support for other 64-bit processors like Itanium or Opteron a relatively simple matter).

78. The entire direction of Linux development changed with IBM's entry into the open-source community and its concerted efforts to control the community for its own economic benefit.

IBM did not change the direction of the Linux development effort. By hopping onboard the Linux bandwagon, it merely amplified trends already in play for ten years before IBM became actively involved. It is SCO that is attempting to hijack and divert the Linux community's effort, not IBM. Representatives of the community supposedly being dominated by IBM have clearly not rallied around SCO as some kind of liberator. Instead, the open source community has unequivocally condemned SCO and sided with IBM.

IBM's first official software application product release for Linux was DB2 in 1998, which was done in response to market leader Oracle announcing Linux support for their entire product line months earlier. IBM eased into Linux support gradually, only becoming involved with Linux kernel development at all upon joining the Trillian project on August 10, 1999, after Red Hat's IPO. Magazine columnists were already predicting, "Linux will wipe out SCO and Unixware" before IBM started doing Kernel development.

In terms of the timeline of IBM's support for Linux, although individual employees of IBM have been Linux supporters, IBM itself ignored Linux for many years. A division of IBM donated some surplus Power PC hardware to Linus in 1994 (back when he was still a graduate student in Finland), but Linus didn't even bother to turn it on for over two years. IBM didn't try again until 1998, with the release of a beta test version of its DB2 database product for Linux, which was the first piece of software IBM ported to Linux. (IBM did not initially offer DB2 for sale on Linux in 1998, instead it gave away unsupported beta-test copies to educational and research institutions.) IBM ported DB2 to Linux a few months after database leader Oracle announced it was porting the entire Oracle product line to Linux; IBM was merely responding to others in the market. IBM's "wait and see" attitude continued for some time.

The Linux development community doesn't take orders from anyone. The Oracle corporation provided the first really big commercial application for Linux, and yet the community's technical judgement can be just as harsh to them as it can to anyone else, as exemplified by this post from Linus Torvalds:

The thing that has always disturbed me about O_DIRECT is that the whole interface is just stupid, and was probably designed by a deranged monkey on some serious mind-controlling substances [*] [footnote: In other words, it's an Oracleism.]...

Sadly, database people don't seem to have any understanding of good taste, and various OS people end up usually just saying "Yes, Mr Oracle, I'll open up any orifice I have for your pleasure".

None of this was at all atypical.

79. In furtherance of its plan to destroy its UNIX competitors, IBM has announced its intention to make Linux, distributed to end users without a fee, the successor to all existing UNIX operating systems used by Fortune 1000 companies and other large companies in the enterprise computing market.

IBM announced plans to push a new product and to increase market share. SCO acts surprised by this, although considering how long it's been since SCO increased its market share, it is perhaps unsurprising that this is a difficult concept for them to grasp.

If IBM had a plan to "destroy its UNIX competitors", it has been wildly unsuccessful. Why would IBM attempt to destroy Hewlett Packard by allowing HP to beat it to market with Linux offerings, or to allow HP to make more money off of Linux in 2001 and 2002 than IBM did?

If IBM is trying to destroy other companies with Linux, why is it petitioning for membership in the new Linux CE Forum, an alliance between Sony, Panasonic, NEC, Royal Philips Electronics, Samsung, Sharp, and Toshiba to push Linux into consumer electronics devices like cell phones, set-top boxes, and the onboard computers in cars. IBM was not a founding member of this forum, and neither UnixWare nor OpenServer has ever been a part of the consumer electronics market.

IBM's competitor Sony has been using Linux as the development environment for its wildly popular Playstation 2 product since its introduction, and last year Sony announced its Cocoon product line of linux-based consumer devices, including personal video recorders and cell phones.

SGI has been a prominent supporter of Linux for years, why would it do that if IBM had bent the project to IBM's proprietary purposes?

Intel is another long-time supporter of Linux. Linus Torvalds just took a leave of absence from Transmeta to work for a year at the Open Source Development Lab. Intel's press release about the opening ceremonies of OSDL does mention IBM in the list of sponsors, along with Intel itself, Hewlett Packard, NEC, Fujitsu, Hitachi, Computer Associates, Mitsubishi Electric, Dell, SGI, Red Hat, SuSE, and others. Caldera itself was one of the founding members of OSDL, and the press release from Intel is dated after Caldera purchased SCO.

IBM was not interested in taking market share or revenue away from SCO, because even at its peak SCO's entire market was beneath IBM's notice. (This is why IBM let AIX for the PS/2 atrophy.) IBM announced its intention to take market share away from Sun and HP, companies that make MORE money than it does selling into the Unix and Linux markets. But IBM has never been particularly excited about taking away SCO's fast-food cash register business.

The IBM PC and its clones were commodity hardware, assembled from interchangeable parts available from multiple vendors, and allowing Fortune 500 corporations and mom-and-pop shops in strip-malls to compete on even terms. Linux does the same thing in software. Linux is commodity software, assembled from interchangeable parts available from multiple vendors, and allowing a level playing field to players large and small.

The lead author of this document, Rob Landley, has assembled his own working Linux distribution, entirely from individual parts downloaded as source code and compiled , following the technical instructions on the Linux From Scratch website. No special equipment was required. The author could go into business competing against Red Hat or SuSE if he really wanted to.

The commodity nature of the software components makes assembling a Linux system so easy that SCO/Tarantella considered launching its own Linux distribution while negotiating to sell its Unix business.

Proprietary Unix competing against commodity open-source Unix is clearly and predictably on the losing end of the battle. No one company can dominate in a commodity market, it is inherently a level playing field. This was true of PC hardware (which drove most other microcomputer vendors like Commodore, Atari, Texas Instruments, and Tandy/RadioShack from the field, consumed the Unix workstation business of companies like Sun, reduced Apple to a relative niche market, and finally wrenched control of PC hardware standards away from IBM itself in the late 1980s under the onslaught of cheap PC clones). Linux is merely the same thing happening on the software side, it's the natural commoditization of a mature market.

80. However, as IBM executives know, a significant flaw of Linux is the inability and/or unwillingness of the Linux process manager, Linus Torvalds, to identify the intellectual property origins of contributed source code that comes in from those many different software developers. If source code is code copied from protected UNIX code, there is no way for Linus Torvalds to identify that fact.

Linus Torvalds has taken issue to this allegation in a recent interview:

"I care deeply about IP (intellectual property) rights. I've personally got more IP rights than the average bear, and as the owner of the copyright in the collective of the Linux kernel, I shepherd even more. It's what I do, every day. I personally manage more valuable IP rights than SCO has ever held, and I take it damn seriously,"

In October 2002, Linus explained at length his policy on how the linux-kernel's GPL license applies to derived works (in the form of binary-only loadable device drivers). As part of his job as linux-kernel maintainer, he consults with lawyers on intellectual property issues.

The Linux kernel developers' policy is to remove any code whose license is not compatible with the GPL. Recently, the linux-kernel developers removed a device driver from the Linux-kernel tree until its license could be clarified. (There was no question the code had been contributed by the copyright holder, but on closer examination the terms of the license it was distributed under were not compatible with the terms on the rest of the kernel, so it was removed until the copyright holder could issue a corrected license.)

Similar license issues have come up on linux-kernel recently with regard to the GNU Free Documentation License and the license on Amtel device drives. In each case, the solution was to either get a corrected license statement from the vendor, or to remove the code.

SCO did not ask to have any code removed, it started out with a lawsuit. SCO has specifically said they don't want to reveal any alleged code overlaps because they believe the Linux developers would "launder" the Linux kernel, I.E. remove the offending code from Linux. SCO is fully aware that Linux kernel development policy is not only to scrutinize contributions, but also to immediately remove any code which is later determined to have unresolved intellectual property issues. SCO has explicitly stated that it has blocked this self-correcting mechanism in the open-source development community, because they do not want the issue resolved that way. That's their stated reason for keeping secret exactly which code in the Linux kernel they have issues with. (This is incompatible with their statments that they have made good faith efforts to resolve the situation short of a lawsuit.)

Their unstated reason for keeping contested code sections secret is they don't want the community tracking down where that code came from, and showing the publicly available sources from which it was derived.

81. As a result, a very significant amount of UNIX protected code is currently found in Linux 2.4.x and Linux 2.5.x releases in violation of SCO's contractual rights and copyrights.

SCO has alledged to have found "hundreds" of lines of infringing code. The 2.4.0 Linux kernel, released January 2001, contained three million, three hundred seventy seven thousand, nine hundred and one lines of code. Assuming the "hundreds" figure were three thousand lines, this would be less than one tenth of one percent of the Linux kernel code.[10]

If the entire operating system based on the Linux kernel is considered, the number grows even worse. Red Hat Linux 7.1, an older representative Linux distribution, has over 30 million lines of code, so SCO is discussing code that is less than 0.01% of the entire system.

The Functional Limitations of Linux Before IBM's Involvement

82. The first versions of Linux evolved through bits and pieces of various contributions by numerous software developers using single processor computers. Virtually none of these software developers and hobbyists had access to enterprise-scale equipment and testing facilities for Linux development. Without access to such equipment, facilities and knowledge of sophisticated development methods learned in many years of UNIX development it would be difficult, if not impossible, for the Linux development community to create a grade of Linux adequate for enterprise use.

Linus Torvalds had an Alpha processor system donated to him by Digital Equipment corporation in 1994. One of the first things Transmeta did when they hired him in 1996 was put him to work upgrading the SMP version of Linux, which Transmeta used extensively. Beowulf systems date back to 1994, and became widely publicized in 1998, predating IBM's first Linux product (DB2).

SCO itself supported the rise of enterprise Linux for years before joining the UnitedLinux effort, which is headed by SuSE. Here's a recording of an hour-long keynote speech given by Dirk Hohndel, CTO of SuSE, in February 2001, entitled, Open Source and Enterprise Computing.

Linux has moved into areas which SCO has not historically participated in, and enterprise computing is only one of them. When was SCO in a cell phone, or in a supercomputing cluster? IBM put Linux in a wristwatch. When has SCO put any of its products in a wristwatch?

One of the larger niches Linux has taken over without the help of IBM is Computer Graphics Imaging (CGI), used to produce computer animation and special effects in Hollywood. Traditionally, this has been a proprietary Unix stronghold dominated by Silicon Graphics (SGI) with its "Irix" version of Unix. Over the past few years, virtually all of Hollywood's CGI modeling, rendering, video capture, editing, compositing, and playback tasks [11] have migrated to Linux, as detailed in the article Linux Invades Hollywood. This is an area where SCO has never been, and never publicly expressed any plans to go, and has been technologically completely incapable of surviving in.

The move is industry-wide, not just a single company. Dreamworks moved from SGI to Hewlett Packard Linux systems, prompting Ed Leonard, Dreamworks Animation Lead Technologist to give a keynote speech at the O'Reilly Open Source convention. (Dreamworks' "Shrek" was rendered entirely on Linux systems.) George Lucas' Industrial Light and Magic moved from SGI to Dell Linux, and Steve Jobs' Pixar switched from Sun Solaris to commodity Intel based Linux systems from RackSpace.

Pixar is an especially important pioneer in the computer graphics industry, producing the first fully computer generated movie, Toy Story, and going on to box office success with Bug's Life, Toy Story II, Monsters Inc., and its most recent offering, "Finding Nemo". Pixar employed Linux guru Bruce Perens for 12 years (1987-1999). During that time, Perens was in charge of the "Debian" Linux distribution, and left Pixar to become HP's "Senior strategist, Linux and open source" until the Compaq acquisition.

Pixar's history with Linux, and Linux's momentum in the industry, helps explain why Pixar moved to Linux instead of MacOS X. Steve Jobs is CEO of both Pixar and Apple; Pixar started when Steve Jobs purchased Industrial Light and Magic's first computer graphics division for $10 million in 1985.

This is a multi-billion dollar industry in which Linux plays a pivotal role, without the help of IBM and without any participation at all from SCO.

83. As long as the Linux development process remained uncoordinated and random, it posed little or no practical threat to SCO or to other UNIX vendors since the original Linux developers did not have access to multiprocessor code or multi-processor development methods needed to achieve high-end enterprise functionality.

The Linux development process was never uncoordinated and random. It was coordinated through the internet, and archives of the constant communication between developers are available dating back to 1991.

As for developers not having access to multiprocessor hardware or enterprise functionality, Caldera itself supplied some, and the first assignment Transmeta gave Linus when they hired him was to upgrade Linux for SMP in 1996. Beowulf clusters go back to 1994, and Linux clusters showed up in the top 500 supercomputers list in 1998 (before IBM released DB2 for Linux) thanks to Nasa and the Lawrence Livermore National Laboratory.

As time went on, SMP systems became cheaper. Truly cheap commodity SMP systems became available retail with the release of dual processor Athlon systems. The first SMP Athlon motherboard was commissioned by Boxx Technology from Tyan. The lead author of this document worked for Boxx at the time, and was the original person lobbying Boxx to produce an Athlon system. (Tyan owed Boxx's designers a favor because we took their inventory of Rambus systems off their hands.) The result was the Tyan Thunder K7 motherboard, as used in the other author's 2001 Ultimate Linux Box, which was the cover story of the November 2001 issue of Linux Journal.

84. To make Linux of necessary quality for use by enterprise customers, it needed to be re-designed and upgraded to accommodate complex multi-processor functionality that has taken UNIX nearly 20 years to achieve. This re-design is not technologically feasible or even possible at the enterprise level without (a) a high degree of design coordination, (b) access to expensive and sophisticated design and testing equipment; (c) access to UNIX code and development methods; (d) UNIX architectural experience; and (e) a very significant financial investment.

In the highlighted phrase, SCO complains that the Linux community only took about nine years to get really world-class SMP support (1991-2000), where it took SCO's predecessors twenty (presumably 1969-1989). It seems like SCO is simply complaining "We're slow, those Linux guys shouldn't code so fast", but actually there's more to it.

First off, AT&T's original Unix started out as the pet project of a group of researchers on obsolete, underpowered, secondhand hardware. It was then adapted to single-processor DEC minicomputers (the PDP-11 and VAX lines), which were not SMP capable (a single VAX processor was the size of a refrigerator). Unix was not even widely used as a 32-bit system until the 1980's (and that was thanks to BSD's VAX version, funded by the Defense Advanced Research Projects Agency, a federal program which also sponsored the early internet).

Beyond that, AT&T was not a computer company; it bought computers from other companies, and it did so on a budget. Unix was used on minicomputers (not mainframes) for its first ten to fifteen years, because mainframes were too expensive to trust to an unusual new operating system. Up in the mainframe world, IBM had its own SMP systems long before it became involved in Unix, and Sequent developed its first NUMA multi-processor system (running BSD) in 1983, only 10 years after Unix was first released outside of Bell Labs.

In 1983, SCO was selling a stripped down version of Unix to run in a 16-bit processor that could access only 1 megabyte of RAM, and booting from floppies because the original IBM PC had no hard drive. None of Intel's x86 processors before the Pentium had any support for multi-processor operation. The Intel Multi-Processor Specification saying how to do it only came out in 1994[12]; the few PC SMP systems before that were improvised, nonstandard, unreliable, and extremely expensive. Multi-processor PC hardware has only became cheap and widely available since 2000, with dual Athlon motherboards and the Hyper-Threading features of the Pentium 4.

The main reason it took Linux as long as it did to support SMP is the developers didn't think it was worth much effort while SMP hardware remained extremely rare. But on June 14, 1994, Alan Cox posted to the linux-kernel mailing list:

Get me a 3+ CPU SMP motherboard and the low level programming docs for the SMP and I have an entire computer society who'll happily do the rest if they get to keep the board.

(More comments from him on that topic may be found here, h ere.) Alan eventually found Caldera, which donated the appropriate hardware, leading to to the addition of SMP in Linux. By early 1997, SMP support was already being upgraded to improve scalability (using multiple locks to improve granularity and parallelism). By 2000, the quality of Linux SMP support had surpassed most commercial offerings.

But SMP is just one example of a larger problem. SCO doesn't know what's necessary to design enterprise functionality because they've never done it. So let's go over the points one letter at a time.

A) the Linux community has a high degree of design coordination through the internet: the linux developers are in constant communication with each other through the linux-kernel mailing list, as well as IRC and in-person meetings at venues like the Linux Kernel Summit at the annual Ottawa Linux Symposium.

B) The only design and testing equipment the linux kernel developers need is the hardware the systems will run on. They are developing software, and the tools they do it with are also software: compilers, debuggers, profilers... all software. (There are web pages for gdb, kdb, valgrind, the Stanford Checker, and smatch. There was some talk recently of adding debugging markers to the Linux kernel itself, but they have so far proven unnecessary.)

C) SCO is using the conclusion it wishes to reach as an assertion, to support itself. In reality, the Linux developers have a multitude of published Unix standards, including Posix and the Single Unix Specification. A fully automated Posix Test Suite was recently released.

In terms of books, the Linux developers had Maurice J. Bach's "The Design of the Unix Operating System", the Minix book, the BSD books and BSD code... Linus Torvalds listed several of the books he'd used in creating Linux back at the start of 1992, see here and here.

On top of that, the thousands of independently developed Unix applications which have been converted to run on Linux serve as the real test harness. If a program fails in a reproducible manner, it's easy to find out why and correct the problem. In the absence of a failing program, technical differences between operating system implementations are academic and irrelevant.

Linux also has tens of thousands of computer science graduate students (like Linus was originally) who are young, eager, energetic, and quite willing to reinvent the light bulb as a learning experience. This brings us to...

D) No experience necessary, they learn on the job. There are thousands of Unix programs available to test out on the new system, and they can improve performance by studying benchmarks. The Mindcraft tests are a good example of this, revealing the "thundering herd" problem. (Something other operating systems had dealt with years earlier, but this was the first time the Linux crowd noticed it.) They introduced wake-one semantics to deal with it, and moved on. This sort of thing happens every week, resulting in a steady increase in performance and scalability. Dan Kegel's Mindcraft Redux page provides a very detailed analysis of Linux scalability improvements that have occurred since Microsoft hired Mindcraft to discredit Linux.

E) If financial investment was all it took, Microsoft Windows would be a technically superior product to any Unix. The enormous proliferation of web pages in the early 1990s did not require financial investment from any central organization owning the internet. Yahoo evolved out of a graduate project of two university students. The current search engine leader, Google, has been running entirely on Linux systems from the start, and employs full-time engineers who contribute improvements back to the Linux kernel because a better Linux kernel helps them be a better search engine. (You can listen to recordings of two speeches from Jim Reese, Chief Operations Engineer of Google: The Technology Behind Google and a related Question and Answer session with Jim Reese.) Thousands of companies contribute their improvements back to Linux for the simple reason that it's too much work to maintain their own custom version, and much cheaper and easier to simply have the fix included in the next release of Red Hat or SuSE, and not have to worry about losing repairs when you upgrade.

That said, IBM announced that starting in 2001 it would spend $1 billion a year on new Linux development. Spending money on new development is not actually illegal, and for a company of IBM's size, $1 billion a year is not a disproportionate committment to a technology; they spend more than that on their PowerPC processor. IBM only made this announcement for 2001: after the founding of Caldera and Red Hat, after Oracle jumped on the Linux bandwagon, after Linus Torvalds was on the cover of Forbes, after "Beowulf" Linux clusters started showing up in the top 500 supercomputer list, after Red Hat and Caldera went public as Linux companies...

IBM's Scheme

85. As market awareness of Linux evolved, IBM initiated a course of conduct with the purpose and effect of using Linux to unfairly compete in the enterprise market. At that point in time, four important events were occurring simultaneously in the enterprise software computing marketplace:

There's a difference between "compete" and "unfairly compete". If IBM was attempting to unfairly compete, it was unsuccessful: HP made more money off of Linux last year than IBM did.

a) Intel chips were becoming widely demanded by enterprise customers since Intel's processing power had increased and its cost had remained low;

This would be x86 chips. Intel was actually losing market share to AMD's Athlons at the time. (The situation has stabilized somewhat since the release of the Pentium 4 and Intel's move away from the widely criticized Rambus technology. Yes, there are people who track this.)

b) SCO's market power in the enterprise marketplace was increasing based on the combined capabilities of SCO OpenServer, SCO UnixWare and SCO's unique position as UNIX on Intel;

Which SCO? SCO/Tarantella was abandoning the Unix market. Caldera was a Linux vendor, although it wasn't very good at it and was losing market share to other vendors like Red Hat and SuSE.

OpenServer, the product that SCO makes most of its money off of today, was a discontinued piece of trash in 1999 that Caldera initially didn't want to buy. The fact it's now the majority of their revenue isn't an endorsement of OpenServer, it means the rest of their business (including their Linux business) dried up and blew away.

And if Caldera is claiming IBM illegally enhanced Linux, yet Caldera can't make money off of Linux despite having been a Linux vendor since 1994 and having released its fastest-selling Linux product ever shortly before the SCO acqusition, while other companies are still making billions off of Linux, then that's even stronger evidence that Caldera's current management is completely and profoundly incompetent.

a) Free Linux had carved a niche in not-for-profit and non-business uses; and

[Editorial note from Rob: No, I didn't screw up the letters here, so don't email me about it. SCO's original complaint goes a) b) a) b). I don't know why.]

The second a) is deceptively misleading. It's true that Linux had a strong showing in not-for-profit and non-business uses, but Linux was already being heavily used in for-profit business uses as well. After all, businesses can make more money when they use a less expensive and better product. Linux has been powering Google single Google's inception. Oracle started supporting Linux in 1998 (and Oracle is NOT a not-for-profit, non-business use; Oracle is to many the definition of "enterprise software").

Linux Weekly News's 1998 timeline shows the early days of Linux's rise to mainstream acceptance, before IBM came on board.

b) IBM was in the process of evolving its business model from products to services.

Actually, IBM was always a services company. Back during the Great Depression of the 1930's, IBM survived because it leased out its tabulator machines instead of selling them, and thus had a steady revenue stream when other companies were going belly-up.[13] A pointy-haired CEO screwed this up in the 1980's (selling off the golden goose for short-term gain by taking its customers off a rental model, and booking huge revenues from one-time sales of the existing deployed equipment). This resulted in the 1992 crash of IBM's stock (the loss of PC leadership didn't help, but that wasn't what sucker-punched their bottom line), and Lou Gerstner put it back on track again when he became CEO after the crash.

Following IBM's return to its service roots, SCO/Caldera was also attempting to become a service company. SCO still has a service division, it just doesn't produce much revenue. SCO/Caldera was unsuccessful at pursuing a strategy IBM started first and executed better, and they're trying to make it sound like IBM did something wrong.

86. In the process of moving from product offerings to services offerings, IBM dramatically increased its staff of systems integrators to 120,000 strong under the marketing brand “IBM Global Services.” By contrast, IBM's largest historic competitor as a seller of UNIX software, Sun Microsystems, has a staff of approximately 12,000 systems integrators. With ten times more services-related personnel than its largest competitor, IBM sought to move the corporate enterprise computing market to a services model based on free software on Intel processors.

This paragraph is basically saying that IBM was much bigger and much more successful than SCO. Does this come as news to anybody?

IBM acted intelligently. The last major wave of commoditization in a maturing computer market was the rise of commodity PCs. IBM was on the wrong side of that one, trying to assemble commodity parts into a proprietary product and losing control of the market to clones of its own design. It was a humiliating lesson which IBM hasn't forgotten.

This time, the now-mature software market is commoditizing, and IBM wants to be on the winning side: providing services, support, and integrated solutions within the level playing field provided by the commodity market.

IBM decided to support Linux because it learned the lesson of the PC. When the market is ready to commoditize, don't stand in its way.

87. By making the Linux operating system free to end users, IBM could undermine and destroy the ability of any of its competitors to charge a fee for distribution of UNIX software in the enterprise market. Thus, IBM, with its army of Global Services integrators who earn money by selling services, would gain a tremendous advantage over all its competitors who earn money by selling UNIX licenses.

IBM didn't make the Linux operating system free to end users. The Linux operating system was already free to end users. Linux has always been a free product, since its inception. You can't blame IBM for Linux being free!

That said, there's a business to be had shipping boxed copies of Linux. SCO's current CEO, Darl McBride, liked to compare selling boxed copies of Linux to selling bottled water. Doing so used to be Caldera's core business, and is still the core business of Red Hat and SuSE. Caldera's Linux was a boxed retail product in 1997. IBM's actions have not undermined Red Hat or SuSE's profitability, they've enhanced them. SCO's failure is SCO's fault.

Taking a second look at paragraph 87: SCO isn't actually claiming that IBM did anything illegal. IBM commoditized their co-factors to undercut the competition, and made a load of money selling more support and services. That's good business. Caldera's tried to do the same thing, that's why they have a services arm. They're just not very good at it.

88. To accomplish the end of transforming the enterprise software market to a services-driven market, IBM set about to deliberately and improperly destroy the economic value of UNIX and particularly the economic value of UNIX on Intel-based processors.

Linux was already undercutting other Unix vendors. The first casualty was the Mark Williams Company (maker of the PC Unix clone Coherent) which went out of business in 1995. Consolidation in the Unix market was predicted by SCO/Tarantella's director of marketing, Greg Schwarzer, in February 1999:

In the future, Schwarzer predicts that the Unix market will consolidate. Monterey, along with Sun Microsystem's Solaris and Hewlett-Packard's HP-UX, will remain as contenders, while others such as Silicon Graphics' Irix, Compaq's Tru64 Unix (formerly Digital Unix), and Data General's DG/UX will fall by the wayside.

Many other industry observers made similar predictions.

89. Among other actions, IBM misappropriated the confidential and proprietary information from SCO in Project Monterey. IBM thereafter misused its access to the UNIX source code.

This is SCO's core assertion. This seems to be the reason they're picking on IBM rather than on some other vendor like HP or Sony. It's buried in paragraph 89 because by itself, it is weak and unsupportable.

The problems with SCO's core assertion include that misappropriated technology was never necessary to make Linux a success, IBM was only one of thousands of companies with access to ancient System V code, and nobody in the industry cares about System V anymore because it's obsolete rubbish, and SCO is accusing IBM of doing things that SCO itself repeatedly announced plans to do. Not to mention the fact that they have been unable to show any misappropriated code snippets that survive scrutiny, and are reduced to Joe "witchhunt" McCarthy's old technique of waving a piece of paper and claiming it proves something without letting anybody else see it.

90. On or about August 17, 2000, IBM and Red Hat Inc. issued a joint press release through M2 Presswire announcing, inter alia, as follows:

IBM today announced a global agreement that enables Red Hat, Inc. to bundle IBM's Linux-based software.

IBM said it would contribute more than 100 printer drivers to the open source community. With these announcements, IBM is making it easier for customers to deploy e-business applications on Linux using a growing selection of hardware and software to meet their needs. The announcements are the latest initiative in IBM's continuing strategy to embrace Linux across its entire product and services portfolio.

Helping build the open standard, IBM has been working closely with the open source community, contributing technologies and resources.

IBM joined the Trillian project (a joint effort with Intel, HP, SGI, and others to port Linux to the Itanium, in direct competition with Monterey) on August 10, 1999, more than a year earlier. Yet the best SCO can come up with is a press release about bundling printer drivers in Red Hat.

91. Thereafter, on December 20, 2000, IBM Vice President Robert LeBlanc disclosed IBM's improper use of confidential and proprietary information learned from Project Monterey to bolster Linux as part of IBM's long term vision, stating:

Project Monterey was actually started before Linux did. When we started the push to Monterey, the notion was to have one common OS for several architectures. The notion actually came through with Linux, which was open source and supported all hardware. We continued with Monterey as an extension of AIX [IBM UNIX] to support high-end hardware. AIX 5 has the best of Monterey. Linux cannot fill that need today, but over time we believe it will. To help out we're making contributions to the open source movement like the journal file system. We can't tell our customers to wait for Linux to grow up.

If Linux had all of the capabilities of AIX, where wecould put the AIX code at runtime on top of Linux, then we would.

Right now the Linux kernel does not support all the capabilities of AIX. We've been working on AIX for 20 years. Linux is still young. We're helping Linux kernel up to that level. We understand where the kernel is. We have a lot of people working now as part of the kernel team. At the end of the day, the customer makes the choice, whether we write for AIX or for Linux.

We're willing to open source any part of AIX that the Linux community considers valuable. We have open-sourced the journal file system, print driver for the Omniprint. AIX is 1.5 million lines of code. If we dump that on the open source community then are people going to understand it? You're better off taking bits and pieces and the expertise that we bring along with it. We have made a conscious decision to keep contributing.

Robert LeBlanc said that IBM was willing. He didn't say that IBM was able. IBM has always respected its intellectual property agreements. It's common practice when open sourcing proprietary codebases, such as Netscape releasing the source code that became Mozilla, or Sun releasing the Star Office code that became Open Office, to audit the code and remove licensed technology the owner of the majority of the codebase does not have authority to release. The resulting code usually won't even run until open-source developers fill in the holes.

The prototype for doing an audited proprietary release was the Netscape source release in 1998, which was announced on January 22, 1998, but did not actually occur until March 31, 1998. As Netscape explained:

Exactly what source code will be available on the Net?

Netscape will make available a developer release of the source code for the 5.0 version of Netscape Communicator Standard Edition, which includes Netscape Navigator, by March 31. Please note that this developer release is intended only for developers, and is not intended to be an end user beta. Netscape will continue to update the source code on the Net as features are added. Netscape will not include code developed by third-parties where Netscape does not have the right to distribute source code (e.g., third-party plugins) or code which it cannot make available due to government regulations (e.g., cryptographic security code).

Can Netscape release something now to help people get started?

Netscape is working as quickly as possible to make preliminary documentation and source code available by March 31, 1998.

Does Netscape plan to also release source code for a Java version of Communicator?

Netscape will only be releasing source code for the 5.0 version of Netscape Communicator Standard Edition. Currently, there are no plans to release source code for a Java version of Communicator.

In March 2001, Linux luminary Bruce Perens, once head of the volunteer-produced "Debian" distribution of Linux, advised his employer HP not to open-source a product called OpenMail, explaining his rationale:

When a company decides to release existing proprietary code as Open Source, the show-stopper is almost always the other parties outside of that company who are involved. Such parties become involved through patents that have been licensed, proprietary code that has been produced by a third party and embedded into the product, and existing contracts relating to the product that have been entered into with customers or other vendors. These sorts of factors complicate the release of every piece of Open Source software I've consulted on at HP so far, no matter what division it comes from.

So, if OpenMail is released as Open Source, we will have to first sanitise it: remove software that is connected with non-disclosure agreements that we entered, patents that we licensed, proprietary code that we bought but can't relicense, and so on. And we must make sure that the result doesn't bring us into violation of contracts we made with customers and vendors, such as the agreements we made with customers when they licensed the OpenMail product. We don't know how big this sanitisation project is yet, if it's bad, it could cost Millions.

To any open-source developer, it is quite obvious what IBM's Robert LeBlanc meant: the release of an audited, sanitized, and probably heavily redacted version of AIX. Even relatively small proprietary projects, like the open-source release of Borland's Inprise database require the codebase to be extensively scrutinized to remove licensed third party intellectual property.

No one would expect IBM's representative to try to fit that obvious and expected detail in an interview sound bite, and if he did the interviewer would probably trim it as superfluous legal boilerplate. Of COURSE IBM would audit the code first. IBM is biologically incapable of going to the bathroom without running the idea by its lawyers first, and coming up with an ISO 9001 documented procedure, which it then follows to the letter, because that's how it thinks. We wouldn't expect anything else. (I'm almost certain IBM has documented intellectual property auditing procedures that predate the first Unix prototype at Bell Labs, let alone the founding of SCO/Tarantella. And I suspect that its modern ISO 9001 procedures for doing it are quite a bit longer.)

92. IBM, however, was not and is not in a position legally to “open source any part of AIX that the Linux community considers valuable.” Rather, IBM is obligated not to open source AIX because it contains SCO's confidential and proprietary UNIX source code, derivative works and methods.

The portions of AIX that contain Unix source code that was not part of BSD or the open-sourced System III Unix release would have to be removed. But since the Berkeley developers managed to do this, it was unlikely to be beyond IBM's abilities.

Also, given that the historical Unix code then belonged to a Linux vendor (Caldera) which was making numerous statements at the time about combining Linux and Unix, and releasing Unix code under open-source licenses (which they did with System III and earlier), IBM had every reason to believe that getting permission to open-source any ancient 1980s source code contained in AIX would be granted if IBM asked for it. The only reason IBM did not ask is because they didn't open-source AIX. They merely stated that they, IBM, had no objection to doing so if the community expressed interest, which the open-source community had not done and still has not done. That is why it never progressed far enough to worry about permissions or intellectual property audit teams and procedures.

As for SCO's interest in Monterey, if that was a concern then a version of AIX released before the start of the Monterey project could have been the basis for any hypothetical release of AIX. It never got as far as selecting a version, because there was no release and no serious movement towards a release.

93. Over time, IBM made a very substantial financial commitment to improperly put SCO's confidential and proprietary information into Linux, the free operating system. On or about May 21, 2001 IBM Vice President Richard Michos, stated in an interview to Independent Newspapers, New Zealand, inter alia:

If IBM had shown any interest in putting SCO's proprietary information into Linux, it could have purchased SCO outright when the stock was trading at under $1/share. IBM doesn't need a five year committment to spend a billion dollars per year to misappropriate SCO's technology. IBM is spending a billion dollars per year on new Linux development. (Why would misappropriating SCO technology require IBM to spend money to do so?)

IBM will put US $1 billion this year into Linux, the free operating system.

IBM wants to be part of the community that makes Linux successful. It has a development team that works on improvements to the Linux kernel, or source code. This includes programmers who work in the company's Linux technology center, working on making the company's technology Linux-compatible.

That team of IBM programmers is improperly extracting and using SCO's UNIX technology from the same building that was previously the UNIX Technology Center.

If SCO is referring to IBM's "900 block" of research buildings in Austin, Texas, they have a number of uses. The OS/2 development team which the lead author of this document (Rob Landley) worked with years ago was in building 903, and the AIX guys (who went on to do JavaOS development) were in building 901. (The OS/2 developers and AIX developers did not speak to each other outside of the cafeteria in building 908, and did not have access to each other's codebases.)

I still have friends who work at IBM, and they tell me that The Linux Technology Center is in the basement of building 908. (Go down the stairs in front of the cafeteria, and instead of going into the library go through the other set of doors.) The people I know who work there are ex-OS/2 developers, not AIX developers. (This could be one reason project Monterey and project Trillian ran in parallel: they were two different departments under two different management chains of command, in two different buildings. This sort of thing is common at IBM, it's a big company.)

The president of SCO/Tarantella understood this when he stated IBM's participation in the Trillian project posed no threat to Monterey on August 20, 1999. He explained:

“IBM has joined so many things over the years, it’s not funny. I don’t think it’s put a significant amount of money into Trillian and it’s a small amount of money to be part of a trend. It’s important to look at internal sponsorship in IBM. Monterey is supported by Bob Stevenson, who’s head of IBM’s server group and Lou Gerstner [IBM’s chief executive] has also bought it off as a strategic initiative.”

He added: “IBM is not looking at Trillian as an alternative to Monterey. The real interest in Linux is coming from all the software companies that sell databases and transaction based tools because they are frustrated that Microsoft more or less gives these things away as part of the Back Office bundle. So they say ‘if you give us a free OS, we’ll make money from it’.”

94. In a news article issued by e-Business Developer on or about August 10, 2001, the following conduct was attributed to IBM regarding participation in the open-source software movement:

Another example is when IBM realized that the open-source operating system (OS) Linux provided an economical and reliable OS for its various hardware platforms. However, IBM needed to make changes to the source to use it on its full range of product offerings.

IBM received help from the open-source community with these changes and in return, released parts of its AIX OS to open source. IBM then sold its mainframes running Linux to Banco Mercantile and Telia Telecommunications, replacing 30 Windows NT boxes and 70 Sun boxes respectively - obviously a win for IBM, which reduced its cost of maintaining a proprietary OS while increasing its developer base. IBM's AIX contributions were integrated into the standard Linux source tree, a win for open source.

Had SCO UnixWare ever previously run on a bank's mainframe? The article talks about porting Linux to an IBM mainframe, not something SCO's involvement would assist in any way.

The article does not say what portions of AIX IBM released. SCO is implying, without evidence, that these were portions IBM did not have the authority to release. Most likely, it was referring to the JFS filesystem, which was a piece of OS/2 technology that IBM released, headed by ex-OS/2 developer Ben Rafanello [14]

95. Again, “IBM's AIX contributions” consisted of the improper extraction, use, and dissemination of SCO'S UNIX source code, derivative works and methods.

No, they didn't. There's a lot more to AIX than just technology licensed from SCO. IBM makes more money from AIX every year than SCO's entire annual gross revenue. SCO's best-selling product is primarily used to power cash registers at McDonald's. SCO considers getting McDonald's to upgrade to a new version of OpenServer to be worth putting out a press release about.

IBM, meanwhile, is selling AIX to banks, which are some of the most demanding enterprise customers there are. And there's a lot more to IBM than AIX, and a lot more to IBM's $1 billion annual budget for new Linux development than dusting off existing technology.

SCO keeps implying that without IBM, Linux was nothing. That IBM made Linux an interesting operating system. So why does IBM offer a redbook on how to port Linux applications to AIX? Why does IBM offer a Linux emulation tool for AIX which allows Linux binaries to run on AIX in a manner similar to SCO's lxrun package for Linux, or the Linux Kernel Personality built into the newest release of UnixWare? Why does IBM offer a precompiled collection of open source Linux applications ported to AIX? (Why has SCO been offering a similar collection of open-source software ported to OpenServer and UnixWare through the SCO Skunkware website for most of a decade now?)

96. In a news article issued by IDC on or about August 14, 2001, the following was reported:

IBM continued its vocal support of the Linux operating system Tuesday, saying the company will gladly drop its own version of UNIX from servers and replace it with Linux if the software matures so that it can handle the most demanding tasks.

IBM executives speaking here at the company's solutions developer conference outlined reasons for the company's Linux support, pointing to features in the operating system that could push it past UNIX for back-end computing. While they admit that Linux still has a way to go before it can compete with the functions available on many flavors of UNIX, IBM officials said that Linux could prove more cost-effective and be a more user-friendly way to manage servers.

'We are happy and comfortable with the idea that Linux can become the successor, not just for AIX, but for all UNIX operating systems,' said Steve Mills, senior vice president and group executive of the IBM Software Group, during a news conference.

Just as the PC became the successor to most other proprietary computer systems. This is the consolidation of a mature industry around a commodity standard. IBM is not responsible for this trend, it's preparing for the inevitable, and ensuring that it is well-positioned when it happens. IBM's muscle couldn't force Windows aside in favor of OS/2, and couldn't displace Solaris with AIX. There is more at work here than IBM, Linux is a force in its own right. Making good business decisions in a changing market is not actionable.

97. Continuing with its “happy and comfortable” idea that Linux succeeds at the expense of UNIX, on or about January 23, 2003, IBM executive Steve Mills gave a keynote speech at LinuxWorld, a trade show, which was reported by Computer Reseller News, IBM's Mills: Linux Will be on Par with UNIX in No Time, January 23, 2003, inter alia, as follows:

IBM will exploit its expertise in AIX to bring Linux up to par with UNIX, an IBM executive said Thursday.

During his keynote at LinuxWorld here, IBM Senior Vice President and group executive Steve Mills acknowledged that Linux lags behind UNIX in scalability, SMP support, fail-over capabilities and reliability—but not for long.

‘The pathway to get there is an eight-lane highway,' Mills said, noting that IBM's deep experience with AIX and its 250-member open-source development team will be applied to make the Linux kernel as strong as that of UNIX. ‘The road to get there is well understood.'

* * *

Mills hinted that the company's full development capabilities will be brought to bear in engineering the Linux kernel to offer vastly improved scalability, reliability and support for mixed workloads—and to obliterate UNIX.

IBM is clearly talking about new development here. They committed a billion dollars per year, an experienced 250 member development team (which rapidly grew much larger, although even 250 developers is over three times as many developers as SCO/Caldera now employs[15]), and corporate experience supplying a similar operating system to banks and governments.

IBM is talking about adapting Linux to the existing customer base to which it was selling mainframe operating systems before Unix was invented. It's talking about adding capabilities to Linux that do not exist in SCO's current product line. SCO does not sell UnixWare to Mastercard and Visa for use in their central credit card transaction processing computers. SCO is not in the same league as IBM, Sun, and HP, and never has been. SCO's best-selling product continues to be OpenServer, and its showcase usage continues to be McDonald's cash registers.

And speaking of speeches given at LinuxWorld, Caldera participated in LinuxWorld long before IBM. Caldera's CEO Ransom Love gave a keynote address at LinuxWorld Expo in 2000 on what it would take for Linux to be adopted across the entire landscape, and on what the SCO acquisition meant to Caldera and the Linux world. If you have realplayer installed, you can watch a recording of his entire speech.

From that same LinuxWorld expo, you can also watch recorded keynote speeches from HP's chief scientist Joel Birnbaum, and Michael Dell, CEO and founder of Dell computers. We all agree that Linux defeated SCO/Tarantella in the market, but it was a fair fight, and Caldera was a Linux vendor both before and after it bought SCO. So why are they picking on IBM? The whole computer industry was involved in Linux; they understand the commoditization of software through the parallel with the PC and the rise of commodity hardware. SCO's previous CEO (Ransom Love) knew about this. If SCO's current CEO (Darl McBride) didn't get it, he was blind. (He certainly wasn't ignorant of it; he at least paid the concept of profiting from a branded commodity lip service before going completely lawsuit-happy.)

98. The only way that the pathway is an “eight-lane highway” for Linux to achieve the scalability, SMP support, fail-over capabilities and reliability of UNIX is by the improper extraction, use, and dissemination of the proprietary and confidential UNIX source code, derivative works and methods. Indeed, UNIX was able to achieve its status as the premiere operating system only after decades of hard work, beginning with the finest computer scientists at AT&T Bell Laboratories, plaintiff's predecessor in interest.

SCO is insisting that the only way IBM could expand Linux to do things that SCO's products did not do, and which IBM's non-Unix mainframe products did do, was to steal SCO technology. This is a risible assertion, if only considering the relative sizes of SCO's and IBM's R&D budgets.

Other manufacturers saw a bright future for Linux years before this 2003 memo. In August 2000, the article Dell looks to keep Sun at bay with Linux quotes CEO Michael Dell:

"The open-source collaborative development model is built to succeed in the Internet age," Dell said in his keynote address at the LinuxWorld Conference and Expo in San Jose, Calif. "It makes much more sense than the proprietary model."

Sun had already found itself forced by market pressures to support Linux back in 1999.

And remember, the original reason Caldera purchased Unix from SCO, as summarized in the title of an interview with Caldera CEO Ransom Love: Caldera CEO says Unix buy will help Linux adoption.

Linux shipments grew 212 percent in 1998, and a February 9, 2000 article, "Linux sales surge past competitors", estimated that Linux already had 4 percent of the desktop operating system market, just behind the macintosh. (SCO didn't even register.) Linux momentum clearly was not attributable primarily to IBM, since this occurred before the articles SCO is quoting.

And don't forget about Beowulf clusters.

99. Based on other published statements, IBM currently has over 7,000 employees involved in the transfer of UNIX knowledge into the Linux business of IBM, Red Hat, Inc. and SuSE Linux AG (the largest European Linux distributor). On information and belief, a large number of the said IBM employees currently working in the transfer of UNIX to Linux have, or have had, access to the UNIX Software Code.

One is moved to wonder why IBM would need 7000 employees to transfer Unix knowledge? Those 7000 employees are doing new development. The sheer scale of IBM's Linux development efforts imply that IBM could recreate the entire history of Unix development (the $1 billion figure of historical investment SCO keeps throwing around) from scratch in a single year.

IBM's Coordination of Linux Development Efforts

100. On information and belief, IBM has knowingly induced, encouraged, and enabled others to distribute proprietary information in an attempt to conceal its own legal liability for such distributions:

What is wrong about this [Linux] distribution, is basically the millions of lines of code that we never have seen. We don't know if there are any patent infringements [in this code] with somebody we don't know. We don't want to take the risk of being sued for a patent infringement. That is why we don't do distributions, and that's why we have distributors. Because distributors are not so much exposed as we are. So that's the basic deal as I understand it.

Karl-Heinz Strassemeyer, IBM The Register, 11/19/2002, www.theregister.co.uk/content/4/28183.html

The IBM executive is talking about patent infringment. This case is not about patent infringement. SCO doesn't have any patents. (What Unix patents there are are still owned by Novell.) Patent infringement is much more difficult to detect than other kinds of intellectual property infringement, because it's possible to infringe a patent you've never heard of: you can never be sure there isn't some patent somewhere that you're infringing among the millions of patents granted annually. Due to this, the entire software industry has evolved a patent defense mechanism of mutually assured destruction, where large software corporations patent everything they can and then cross-license their entire patent portfolio with other companies.

SCO is taking a quote from an IBM executive who is concerned about patent infringement, explaining why IBM prefers to outsource some tasks to avoid potential legal liability, and is using that in a completely different context to make allegations about misappropriation of trade secrets.

101. IBM is affirmatively taking steps to destroy all value of UNIX by improperly extracting and using the confidential and proprietary information it acquired from UNIX and dumping that information into the open-source community. As part of this effort, IBM has heavily invested in the following projects to further eliminate the viability of UNIX:

The value of proprietary Unix is being destroyed by commoditization and obsolescence. IBM has chosen the obvious winner in the commodity vs proprietary struggle, the side that SCO/Caldera itself was on from its founding until the start of this year.

a) The Linux Technology Center was launched in 2001 with the advertised intent and foreseeable purpose of transferring and otherwise disposing of all or part of UNIX, including its derivative works, modifications and methods, into an open-source Linux environment;

The Linux Technology Center is IBM's vehicle for doing Linux development. IBM has internal procedures in place, “Chinese Walls” there to prevent the inappropriate transfer of information between divisions. Such a transfer is entirely unnecessary for a company spending a billion dollars annually and seven thousand full-time developers to bring Linux up to the same level of scalability as IBM's existing mainframe operating systems such as OS/390, which is a level of scalability that SCO's own products have never even approached.

b) The IBM Linux Center of Competency was launched to assist and train financial services companies in an accelerated transfer of UNIX to Linux with the advertised intent and foreseeable purpose of transferring and otherwise disposing of all or part of UNIX, including its derivative works, modifications and methods into open source.

IBM is migrating customers from Unix to Linux. SCO's wild assertions about IBM performing entirely unmotivated intellectual property violations continue to be unfounded. Even if SCO had something IBM wanted, which it didn't, IBM could easily reproduce it by in-house development using virgin programmers [16], which IBM has in abundance.

c) A carrier-grade Linux project has been undertaken to use UNIX source code, derivative works, modifications and methods for the unlawful purpose of transforming Linux into an enterprise-hardened operating system;

The Carrier Grade Linux project is coordinated through OSDL, not through IBM. All development of Carrier Grade Linux is public, and the source code and requirements documents are available on the internet.

The charter of carrier grade Linux is "Research and development in making Linux the OS of choice for systems in the telecommunications industry". This is not a market niche in which SCO's existing products are widely used, because SCO's products simply aren't good enough.

d) A data center Linux project has been undertaken to use UNIX source code, derivative works, modifications and methods for the unlawful purpose of transforming Linux into an enterprise-hardened operating system; and

Again, Data Center Linux is a project of OSDL, an organization which includes Intel and Sun and HP and a bunch of other companies. Once again, development is done out in the open and if SCO wanted to point to anything specific rather than complaining that the market has moved on and rendered their products obsolete, it is free to do so.

IBM is not the prime mover in many of these high-end Linux scalability efforts. An article from February 2003 states, "HP said it has been working with the Linux community for years and is trying to help boost the operating system's ability to work on large, multiprocessor servers as a way to increase its Itanium server performance, according to Judy Chavis, worldwide Linux director at HP." The article goes on to quote Ms. Chavis as saying, "We were ahead of IBM for years before they realised Linux was important...". And remember, HP made more money from Linux than IBM did in 2002. (That article has another quote from Judy Chavis of HP: "Linux is a key corporate focus for HP. Every single business unit, whether it be hardware or software, has a Linux roadmap and Linux deliverables.") You can also watch or listen to a january 2002 speech from Hewlett-Packard CEO Carly Fiorina entitled, Fueling Innovation and Opportunity With Linux. SCO's focus on IBM is absurd.

And once again, SCO/Tarantella is welcome to point to any datacenters built around their own products, which mostly run things like fast food cash registers. Corporate datacenters have mainframes or rackmounted computer equipment in climate-controlled rooms with raised floors, not a PC on a shelf above the french fry boiler. Datacenters are totally different, here's a description of the data center IBM set up for Nordstrom's. Many Fortune 500 data centers cost more than SCO's entire gross annual revenue.

e) Other projects and initiatives have been undertaken or supported that further evidence the improper motive and means exercised by IBM in its efforts to eliminate UNIX and replace it with free Linux.

Replacing Unix with Linux is not an improper motive, and the means involves a billion dollars and 7000 full-time developers from IBM doing new development.

102. But for IBM's coordination of the development of enterprise Linux, and the misappropriation of UNIX to accomplish that objective, the Linux development community would not have timely developed enterprise quality software or customer support necessary for widespread use in the enterprise market.

First of all, OSDL is coordinating the development of enterprise Linux, not IBM. (Linus Torvalds recently took a year's leave of absence from transmeta to work at OSDL full-time.) OSDL is funded by many other companies, including Intel and HP.

Secondly, it's ludicrous to think that IBM can't do enterprise development, on any platform, without the help of a tiny little company like SCO. IBM has been the leading computer manufacturer for half a century, and invented many of the enterprise techniques at issue. If IBM can scale mainframe technology down to create PC operating systems like OS/2, it is quite capable of beefing up Linux without reference to AIX, especially if it budgets a billion dollars annually and thousands of developers around the world to do so.

Third, Caldera itself was focused on increasing Linux's enterprise scalability, before the SCO acquisition. Here's a February 24, 1999 press release from Caldera announcing the First Linux Integration of Raid Enterprise Storage. Here's a March 22, 1999 press release about Caldera lining up a round the clock support center, which includes the quote, "We wanted the kind of support that would enable VARs and System Administrators to implement real business-critical, OpenLinux solutions on an enterprise level." Here's a July 15, 1999 press release about a partnership with Computer Associates to increase corporate IT adoption of Linux, and on August 9, 1999 they ported OpenLinux 2.2 to Sun Microsystems' 64-bit Ultrasparc systems.

The first paragraph from this press release reads:

OREM, UT — September 29, 1999 - Caldera Systems, Inc., the "Linux for Business" leader, today announced that it will form a strategic relationship with Fujitsu Ltd. that will include the distribution of Caldera's OpenLinux 2.3 (and future versions) on many of Fujitsu's servers. The move signals Fujitsu's strong commitment to Linux in Japan.

It also notes that at the time, Fujitsu's consolidated annual revenue was $43.3 billion dollars.

There seems to be significantly more evidence of enterprise interest in the version of Caldera OpenLinux released in 1999 (before Caldera purchased SCO's server division) than in the most recent versions of SCO UnixWare and SCO OpenServer combined.

Linux was already developing enterprise quality software before IBM jumped on board. According to a February 19, 1999 Forbes article, Hewlett-Pacard, Dell, and Oracle had all jumped on Linux before IBM shipped its first copy of the operating system. Almost a year earlier, Linus Torvalds had been on the cover of the August 10, 1998 issue of Forbes. Linux shipments increased 212 percent in 1998.

In terms of SMP support, Compaq's engineers made Linux boot on a 31 processor system with 256 gigabytes of memory in September 2000 (Technical announcement here, or Slashdot story here). Meanwhile, SCO's top of the line product, UnixWare 7.1.3 (released December 4, 2002) can only support 8 processors and 32 gigabytes of memory. So three years ago, Linux scaled to four times as many processors and eight times as much memory as UnixWare does today.

By 1998, Linux-based systems were listed in the Top 500 supercomputers list. A Linux system at the Lawrence Livermore National Laboratory (which was not manufactured or installed by IBM) is currently the third fastest supercomputer in the world, and a new Linux-based system being installed at Sandia National Laboratory by supercomputer manufacturer Cray and Linux vendor SuSE (again, not by IBM) is expected to become the fastest supercomputer in the United States. (SuSE is the head of the UnitedLinux effort that SCO/Caldera was a founding member of, and at the start of the lawsuit SCO owned a 25% stake in UnitedLinux. Many of the Linux developers SCO/Caldera laid off over the past few years now work for SuSE and UnitedLinux.)

You can view or listen to a January 2002 talk from Thomas Sterling, of the California Institute of Technology and the NASA Jet Propulsion Laboratory, entitled "Towards Petaflops Computing with Linux". (A Petaflop is one million Gigaflops. A Gigaflop is a billion FLoating Point Operations Per second.)

SCO products do not run on supercomputers at all. SCO only ships products for the hardware found in desktop PC systems. Leaving aside everyting else, SCO simply had no useful technology for IBM to misappropriate in the enterprise, datacenter, or supercomputer space. SCO does not participate in any of those markets, and never has.


(Breach of IBM Software Agreement)

103. Plaintiff incorporates and re-alleges paragraphs No. 1-102, above.

104. As set forth above, SCO is the successor to AT&T under that certain Software Agreement originally executed by and between AT&T and IBM designated as SOFT-00015. The Software Agreement specifies the terms and conditions for use of UNIX System V source code by IBM.

105. With respect to the scope of rights granted for use of the System V source code under Section 2.01 of the Software Agreement, IBM received the following:

[A] personal, nontransferable and nonexclusive right to use in the United States each Software Product identified in the one or more Supplements hereto, solely for Licensee's own internal business purposes and solely on or in conjunction with Designated CPUs for such Software Product. Such right to use includes the right to modify such Software Product and to prepare derivative works based on such Software product, provided the resulting materials are treated hereunder as part of the original Software Product. [Emphasis added.]

106. IBM has violated its grant of rights under §2.01 of the Software Agreement by, inter alia, modifying and assisting others to modify the Software Products (including System V source code, derivative works and methods based thereon) for purposes other than IBM's own internal business purposes. By actively supporting, assisting and promoting the transfer from UNIX to Linux, and using its access to UNIX technology to accomplish this objective, IBM is (a) using the Software Product for external business purposes, which include use for the benefit of Linus Torvalds, the general Linux community and IBM's Linux distribution partners, Red Hat, Inc. and SuSE Linux AG and its subsidiaries; and is (b) directly and indirectly preparing unauthorized derivative works based on the Software Product and unauthorized modifications thereto in violation of §2.01 of the Software Agreement.

107. IBM agreed in §2.05 of the Software Agreement to the following restrictions on use of the Software Product (including System V source code, derivative works and methods based thereon):

No right is granted by this Agreement for the use of Software Products directly for others, or for any use of Software Products by others.

108. IBM has breached §2.05 of the Software Agreement by, inter alia, actively promoting and allowing use of the Software Products and development methods related thereto in an open and hostile attempt to destroy the entire economic value of the Software Products and plaintiff's rights to protect the proprietary nature of the Software Products. By way of example and not limitation, IBM has used protected UNIX methods for others in accelerating development of the 2.4.x kernel and 2.5.x Linux kernel in, among others, the following areas: (a) scalability improvements, (b) performance measurement and improvements, (c) serviceability and error logging improvements, (d) NUMA scheduler and other scheduler improvements, (e) Linux PPC 32- and 64-bit support, (f) AIX Journaling File System, (g) enterprise volume management system to other Linux components, (h) clusters and cluster installation, including distributed lock manager and other lock management technologies, (i) threading, (j) general systems management functions, and (k) other areas. But for the use by IBM of these protected UNIX methods in Linux development, the Linux 2.4.x kernel and 2.5.x kernel capacity to perform high-end enterprise computing functions would be severely limited.

IBM technology has out-scaled SCO technology for many years. (The same is true for Sun, HP, and others.) The original Unix development at AT&T was done on Digital Equipment Corporation minicomputers in the PDP and VAX line, which were significantly less powerful than a modern laptop computer. With the power of computer hardware doubling every 18 months, computer companies must constantly invest in new research and development to stay current. The scalability work to perform well on the new hardware that came out in the late 1980s and early 1990s was not done by AT&T or its successors, but by other companies like Sun, IBM, HP, Sequent, SGI. SCO is claiming to own work that others did to enhance a codebase it inherited, and further is attempting to bar two of those others (IBM and Sequent) from using their own work in other contexts.

Going through the letters, it seems fairly obvious that SCO has read the 2.5 merge candidate list that the lead author of this commentary, Rob Landley, prepared in October 2002. This list eventually contained around 30 projects that were submitted for inclusion in the 2.5 Linux kernel before the first "Feature freeze" in preparation for the release of the next stable version (2.6.0).

It's interesting to note that technology IBM has submitted to Linux has been repeatedly turned down. IBM's Enterprise Volume Management System was rejected, prompting IBM to gracefully withdraw it, and design a new system built on top of the winner (Sistina's LVM2, see the 2.5 merge candidate list for links). Linux luminary Alan Cox (maintainer of the 2.2 kernel series and Red Hat's kernel maintainer for many years) replied, "Throwing away a big piece of code really sucks. I think however its the right path... I appreciate you guys doing what looks to be the right thing for the Linux project overall even when it must be a bitter disappointment." Linux Weekly News wrote an excellent article praising IBM's EVMS move.

IBM was similarly gracious about discontinuing its Next Generation Posix Threading library, in favor of Red Hat's competing design. The respect IBM has earned in the open-source community is precisely due to the fact that IBM has NOT attempted to exert undue control, has gracefully accepted defeat when technical decisions went against it, and has deferred to the established development community wherever possible. IBM has been a good citizen of the open-source world.

Let's go through the bullet points anyway.

"(a) scalability improvements" is too vague to be meaningful, but here's a technical article about how William Lee Irwin improved Linux on x86 machines with 64 gigabytes of memory by the addition of a technology called "page clustering" (which SCO Unixes don't do), and details the original development process he undertook at length. Other scalability technologies SCO is unlikely to have that have recently been publicly discussed and prototyped by Linux developers include swap prefetching, kernel preemption (sponsored by MontaVista), and shared page tables. Audio recordings of talks on many of these subjects and more are available from the Ottawa Linux Symposium website, and some less-technical recordings from LinuxWorld Expo and Atlanta Linux Showcase.

"(b) performance measurement and improvements,"... SCO itself released the AIM Benchmark Suites VII and IX under the GPL, pointing a finger at IBM for this sort of thing is ludicrous.

Beyond that, is SCO really claiming to own the concept of the "benchmark"? People were benchmarking DOS on the original IBM PC in the mid 1980s using benchmarks from Byte magazine. The Linux Benchmark Suite page collects together over two dozen different benchmark suites for performance measurement of operating systems. The Linux Benchmark page is a project run by Dr. Randy Appleton at Northern Michigan University; the project's conclusions make for good reading.

"(c) serviceability and error logging improvements,"... SCO is claiming to own the concept of error reporting.

"(d) NUMA scheduler and other scheduler improvements,"... Several people contributed to the NUMA scheduler in Linux 2.5. Much of the basic work was done by Erich Focht of NEC in Germany. An IBM developer offered a smaller and simpler version, and described the relationship between his version and Erich's version here.

The Sequent developent team IBM had purchased benchmarked the various NUMA patches for Linux against each other, to help decide between the various approaches. Sequent has produced NUMA hardware since the mid 1980's. SCO has never produced its own hardware. SCO is denying credit for programming hardware to the company that produced that hardware.

"(e) Linux PPC 32- and 64-bit support,"... Again, IBM invented the Power PC hardware. IBM manufactuers it. SCO does not ship products for it. SCO is claiming credit away from IBM for supporting Linux on hardware IBM invented. This is completely ludicrous.

"(f) AIX Journaling File System,"... Was created by IBM, not by SCO or any of its predecessors.

A filesystem organizes the data on a computer disk. (Examples of different filesystems would be PC formatted floppy diskette vs a Macintosh formatted floppy diskette.) JFS, which stands for Journaling FileSystem, was independently developed by IBM, and is not a technology they licensed from SCO. IBM independently developed multiple compatible versions of JFS for their different operating system products, including separate (but compatible) versions for AIX and OS/2. The version of JFS merged into Linux was based on the OS/2 version, not the AIX version.

A journaling filesystem is a way of storing data that requires less clean-up work to recover from a power failure, hardware failure, or other type of system crash that doesn't shut down the computer cleanly. IBM's JFS is just one of many types of journaling filesystem, several companies have donated their journaling technology to Linux. Jon "Maddog" Hall, president of Linux International, mentioned the issue of journaling filesystems in a 2000 interview, suggesting more versions of journaling technology were being donated than Linux could actually use:

Let's choose journaling filesystems as an example. There are four or five different companies that have general filesystem technology that they'd like to donate. Well, gang, why don't you all come together and amongst yourselves, make the decision as to which one is going to go in? If it wasn't your technology, you know, grin and bear it. Take the defeat gracefully. It doesn't make sense to have five different journaled filesystems in Linux. So, it's a little bit of the culture that they have to understand.

JFS is one of four different journaling filesystems (Red Hat's ext3, Namesys's Reiserfs, SGI's XFS, and IBM's JFS) currently merged into mainstream versions of Linux. The most popular journaling Linux filesystems are ext3 (on Red Hat systems and derivatives) and Reiserfs (for SuSE systems, including UnitedLinux). The main reason to use JFS is to read data off of existing JFS-formatted hard drives without having to reinstall, I.E. to ease IBM customer migration from OS/2 or AIX to Linux.

"(g) enterprise volume management system to other Linux components,"... SCO is so out of it it didn't notice that IBM's EVMS was rejected by the Linux kernel developers, in favor of LVM2 from Sistina (as mentioned earlier in this paragraph's reply). How does SCO itself do volume management? What is SCO's competing product that IBM is alledged to have stolen technology from?

"(h) clusters and cluster installation,"... According to the Beowulf Project at Goddard Space Flight Center:

The Beowulf Project began at the Goddard Space Flight Center (GSFC) in the summer of 1994 with the assembly of a 16 node cluster developed for the Earth and space sciences project (ESS) by Thomas Sterling and Donald Becker, formerly of USRA/CESDIS. The project quickly spread to other NASA sites, other R&D labs and to universities around the world. The project's scope and the number of Beowulf installations grew at a fantastic rate during the late 1990's. As of mid 2000, the project continues to grow and it scope has expanded. The Beowulf community now influences almost every aspect of high performance computing and in many cases plays a direct leadership role.

As for installation, check out Caldera's Lizard installation tool for Linux, although the leading company for Beowulf cluster management is generally considered to be the one founded by the guy who invented Beowulf clusers, Donald Becker's Scyld Computing Corporation.

"distributed lock manager and other lock management technologies,"... The distributed lock manager is part of NFS, the Network File System invented (and open-sourced) by Sun Microsystems in the 1980's, and standardized by the Internet Engineering Task Force as RFC 1094 in 1987. As usual, SCO is claiming credit for somebody else's technology. Caldera did do some bug fixes on NFS years ago, but they went directly in to the Linux kernel.

"(i) threading,"... SCO didn't invent threading. Sun claims to have brought threading to Unix in 1992 with the introduction of Solaris. (See Sun's history of threading, starting on page 9 of that link.) The Posix specification standardizing threading for Unix systems was issued in 1995 (long before IBM had anything to do with Linux).

But threading existed long before Solaris brought it to Unix: OS/2 1.0 supported multithreading when it was introduced in 1987. So SCO is claiming that IBM can't handle a technology that IBM shipped the first PC implementation of 15 years ago. Neither AT&T Unix nor SCO/Microsoft Xenix supported threading when OS/2 1.0 came out, yet SCO claims it as a trade secret.

Ignorance of what other people have done is not a good basis for a trade secret claim. ("I didn't know you pioneered this decades ago" is particularly weak.)

In addition, we mentioned earlier that IBM's NGPT approach to threading was rejected in favor of Red Hat's NPTL. IBM came up with a good design, but the open-source community came up with a better one. IBM simply is not in the driver's seat here.

"(j) general systems management functions,"... Does SCO perhaps mean something like the Caldera Open Administration System which they released for Linux (open source) in 1998? Or perhaps they mean the Webmin and OpenWBEM projects, which they also take credit for.

Aside from all that, can SCO possibly believe ANY computer company, let alone the world's largest, couldn't manage "general systems management functions" without it?

"and (k) other areas."... This means nothing. (Big surprise.)

109. IBM agreed in §7.10 of the Software Agreement to the following restrictions on transfer of the Software Product, including AIX as a derivative work of UNIX System V:

[N]othing in this Agreement grants to Licensee the right to sell, lease or otherwise transfer or dispose of a Software Product in whole or in part.

In the response to paragraphs 2 and 6 of this complaint, reference was made to Exhibit C. Paragraph 2 of exhibit C explicitly allows IBM to create derivative works that belong to IBM. Paragraph 9 says that IBM is free to use "ideas, concepts, known-how or techniques relating to data processing embodied in SOFTWARE PRODUCTS subject to this agreement", just not to violate SCO's copyrights by including actual code snippets.

110. IBM has breached §7.10 of the Software Agreement by, inter alia, transferring portions of the Software Product (including System V source code, derivative works and methods based thereon), including but not limited to the AIX Journaling File System and all other UNIX-based source code publicly announced by IBM, to Linus Torvalds for open distribution to the general public under a software license that destroys the proprietary and confidential nature of the Software Products.

SCO has no hold over IBM's "derivative works or methods based thereon". Exhibit C, paragraphs 2 and 9 again.

And SCO should really stop acting so shocked at the concept of Linus Torvalds receiving code to be distributed under the GPL. This is not something new to Caldera, Linux was their core business from their founding up until this year, and they attached a copy of the GPL as exhibit 10.14 to their S-1 SEC filing for their initial public offering. SCO has a page devoted to its contributions to Linux.

111. IBM has further stated its intention to transfer the entirety of AIX into open source in anticipatory violation of its obligations under §7.10 of the Software Agreement.

IBM stated it was willing to open-source as much of AIX as it could do so, if the Linux developers were interested (which they weren't). IBM has never released existing proprietary code without an intellectual property audit. One of IBM's engineers who went on to law school wrote "The Intellectual Property Audit". This is deeply ingrained in IBM's culture. SCO is being as insulting to IBM by claiming they'd fail to track intellectual property as it is to the Linux community by claiming we can't code circles around them in our sleep.

112. IBM agreed in Side Letter §9, a substitute provision to §7.06(a) of the Software Agreement, to the following restrictions on confidentiality of the Software Product, including AIX as a derivative work of UNIX System V:

Licensee agrees that it shall hold Software Products subject to this Agreement in confidence for AT&T. Licensee further agrees that it shall not make any disclosure of such Software Products to anyone, except to employees of Licensee to whom such disclosure is necessary to the use for which rights are granted hereunder. Licensee shall appropriately notify each employee to whom any such disclosure is made that such disclosure is made in confidence and shall be kept in confidence by such employee.

113. In recognition of SCO's right of confidentiality of the Software Products, IBM directs all customers who need to view AIX source code to first obtain a “read only” source code license from SCO as a condition to viewing any part of the AIX source code. For example, SCO received a letter on or about March 4, 2003 from Lockheed Martin Corporation requesting verification of the existence of a Software Agreement by and between Lockheed and SCO as a condition to Lockheed obtaining access to view AIX source code. The letter stated, in part, as follows:

LMATM is in the process of licensing [AIX] from IBM to be used for integration purposes only. Per the attached supplement to the subject document, contained within the AIX source code is third party IP which must be licensed from the owner prior to IBM providing the AIX source code to any licensee (see Prerequisite Source Licenses, Para.2.2).

* * *

2.2 Prerequisite Source License. IBM cannot disclose (includes viewing) certain Third-Party Source Code to any party who does not have a license that permits access to the Code. Prior to receiving or accessing the Source Code described above in this Supplement, LMATM must obtain the following Source Code Licenses:

a) AT&T Technologies, Inc., AT&T Information Systems, Inc., or UNIX™ Systems Laboratory Software Agreement No. SOFT—-and AT&T Information Systems, Inc. Software Agreement Supplement for Software Product AT&T UNIX System V Release 4.0, or AT&T Information Systems, Inc. Schedule for Upgrades (from UNIX System V Release 3.1 to UNIX System V Release 3.2 or from UNIX System V Release 3.1 International Edition to UNIX System V Release 3.2 International Edition) or equivalent SCO Group License.

SCO is pointing out that IBM is following the letter of the law with its licensing agreements, whenever any of its licensees want to license AIX source code. One wonders in what respect SCO's lawyers can possibly think this helps SCO's case.

114. IBM has breached its obligation of confidentiality by contributing portions of the Software Product (including System V source code, derivative works and methods based thereon) to open-source development of Linux and by using UNIX development methods in making modifications to Linux 2.4.x and 2.5.x, which are in material part, unauthorized derivative works of the Software Product. These include, among others, (a) scalability improvements, (b) performance measurement and improvements, (c) serviceability and error logging improvements, (d) NUMA scheduler and other scheduler improvements, (e) Linux PPC 32- and 64-bit support, (f) AIX Journaling File System, (g) enterprise volume management system to other Linux components, (h) clusters and cluster installation, including distributed lock manager and other lock management technologies, (i) threading, (j) general systems management functions, and (k) others.

SCO is repeating itself, see paragraph 108 for the rebuttal to this.

115. IBM has further stated its intention to transfer the entirety of AIX into open source in anticipatory violation of its obligations under §7.06 (a) of the Software Agreement.

116. Export of UNIX technology is controlled by the United States government. Thus, SCO, IBM and all other UNIX vendors are subject to strict export control regulations with respect to any UNIX-based customer distribution. To this end, IBM agreed in §4.01 of the Software Agreement to restrictions on export of the Software Product (including System V source code, derivative works and methods based thereon), as follows:

Licensee agrees that it will not, without the prior written consent of AT&T, export, directly or indirectly, Software Products covered by this Agreement to any country outside of the United States.

This provision was later modified to allow export rights to several countries outside the United States. However, no permission has ever been granted by SCO or its predecessors to IBM to allow it to indirectly make available all or portions of the Software Product to countries outside the United States that are subject to strict technology export control by the United States government: viz., Cuba, Iran, Syria, North Korea and Libya. IBM is ignoring and attempting to circumvent the export control restrictions that apply to UNIX as it accelerates development of Linux for enterprise use.

SCO partnered with a German company to produce UnitedLinux, and shipped its Linux development team overseas. If SCO is going to rattle a sword about export restrictions on anything, SCO itself is in deep trouble.

117. Thus, IBM has breached §4.01 of the Software Agreement by, inter alia , making extensive, advanced multiprocessor scaling functions of the Software Product, including derivative works and methods based thereon, available for free distribution to anyone in the world with a computer. As it relates to Linux 2.4.x and 2.5.x releases, IBM is indirectly making the Software Product and operating system modifications available to countries and organizations in those countries for scaling single processor computers into multi-processor supercomputers that can be used for encryption, scientific research and weapons research.

Leaving aside that Compaq got a 32-processor Linux system up in 2000 without IBM's help, I'd like to point out that Caldera takes credit for (and I quote): "Early support of the SMP development effort (hardware provided to the SMP development team)". This agrees with Alan Cox's old Linux SMP page for the 2.0 Linux kernel, which says that the initial SMP support was made possible by Caldera.

118. SCO has the self-executing contractual right to terminate IBM's right to use and distribute the Software Product, including derivative works and methods based thereon, if IBM fails to fulfill one or more of its obligations under the Software Agreement. This authority is contractually granted under the following provisions of the IBM Related Agreements:

If Licensee fails to fulfill one or more of its obligations under this Agreement, AT&T may, upon its election and in addition to any other remedies that it may have, at any time terminate all the rights granted by it hereunder by not less than two (2) months' written notice to Licensee specifying any such breach, unless within the period of such notice all breaches specified therein shall have been remedied; upon such termination Licensee shall immediately discontinue use of and return or destroy all copies of Software Products subject to this Agreement. [Software Agreement, §6.03]

Regarding Section 6.03 of the Software Agreement and Sections 2.07 and 3.03 of the Sublicensing Agreement, we will not terminate your rights for breach, nor will we give notice of termination under such Sections, for breaches we consider to be immaterial. We agree to lengthen the notice period referenced in such Sections from two (2) months to one hundred (100) days. If a breach occurs that causes us to give notice of termination, you may remedy the breach to avoid termination if you are willing and able to do so. In the event that a notice of termination is given to you under either of such Sections and you are making reasonable efforts to remedy the breach but you are unable to complete the remedy in the specified notice period, we will not unreasonably withhold our approval of a request by you for reasonable extension of such period. We will also consider a reasonable extension under Section 2.07 of the Sublicensing Agreement in the case of a Distributor who is making reasonable efforts to remedy a breach.

In any event our respective representatives will exert their mutual good faith best efforts to resolve any alleged breach short of termination. [Side Letter, § 5]

It's too bad that IBM didn't insist on an impartial arbiter to judge what is and isn't immaterial. On the evidence, it appears that the highlighted promise would only be protection if SCO hadn't sold its integrity when it ran out of money.

119. Consistent with these rights, on March 6, 2003, plaintiff delivered a notice of termination to Sam Palmisano, Chief Executive Officer of IBM (the “AIX Termination Notice”) for IBM's breaches of the Software (and Sublicensing) Agreement by IBM. 120. Following delivery of the AIX Termination Notice, plaintiff took every reasonable step to meet and confer with IBM regarding IBM's breach of the Software Agreement and Related Agreements.

121. IBM has disregarded SCO's rights under the AT&T / IBM Agreement by failing to undertake any efforts to cure its numerous and flagrant violations thereunder. As a result, effective June 13, 2003, the AT&T / IBM UNIX Agreement is terminated and IBM has no further rights thereunder.

122. IBM nonetheless continues to operate under the AT&T / IBM Agreement, and use the Software Products and Source Code thereunder as though its rights under the Agreement have not been terminated.

123. IBM no longer has any right to use the UNIX Software Code or make modifications or derivative works thereunder. In fact, IBM is contractually obligated to “immediately discontinue use of and return or destroy all copies of Software Products subject to this Agreement.”

SCO has not proven that IBM is actually in breach of contract, SCO filed a lawsuit before attempting to terminate IBM's contract, and a proportional response to the discovery of any disputed code in the Linux kernel would be to ask for that code to be removed. SCO has not asked for any specific code to be removed, in fact it refuses to reveal specifically which code it considers to be at issue exactly because it believes the Linux developers would remove the code at issue from the Linux Kernel (whether it was actually infringing or not), and SCO does not want the code removed.

124. As a result of IBM's breaches before termination, SCO has been damaged in the marketplace for violations by IBM in an amount to be proven at trial, but not less than $1 billion.

SCO/Caldera bought the Santa Cruz Operation's Linux business for $36 million. SCO/Tarantella bought UnixWare from Novell in 1995 for an estimated $72.4 million[17]. How can SCO possibly have suffered $1 billion worth of damages on something it only cost $36 million to buy?

125. In addition, and to the extent that IBM continues to completely repudiate its obligations regarding the Software Product, plaintiff will sustain substantial continuing and ongoing damages. SCO is entitled to damages in an amount measured by the benefits conferred upon IBM by its ongoing, improper use of the Software Products. These damages include the full amount IBM receives as a result of its ongoing sales of AIX, including software, services and hardware.

IBM has invested billions of dollars in AIX development over the years, and has a license that will never require it to pay another penny of royalties to SCO. Those facts are not in dispute here. SCO is not entitled to any AIX revenue, any Solaris revenue, any HP-UX revenue, or any other version of Unix other than its own and the 5% agent fee it collects for passing on royalties from Novell's old Unix customers.

126. Moreover, if IBM does not return or destroy all source and binary copies of the Software Products and/or continues to contribute some or all of these protected materials to open source, SCO will be irreparably harmed. As a result, SCO is entitled to a permanent injunction requiring IBM to return or destroy all source code and binary copies of the Software Products and/or prohibiting IBM from further contributions of the protected Software Products into open source.

All harm SCO has suffered has been due to market forces. This time last year, SCO/Caldera was a Linux vendor. The decline in SCO's market share, SCO's stock price, and SCO's revenue has not been due to IBM. How can IBM continuing to sell AIX, which IBM has been doing peacefully for a decade and a half, have any affect on SCO?

SCO's contention that System V code is in the Linux Kernel has nothing to do with AIX. AIX is supposed to have System V in it. How would a recall of AIX address alleged code contamination in Linux?


(Breach of IBM Sublicensing Agreement)

127. Plaintiff incorporates and re-alleges paragraphs No. 1-126, above.

128. As set forth above, SCO is the successor to AT&T under that certain Sublicensing Agreement originally executed by and between AT&T and IBM designated as SUB-00015A. The Sublicensing Agreement grants the right to distribute object-based code of UNIX System V and modifications thereto and derivative works based thereon.

129. SCO has terminated IBM's right to use and distribute the Software Product, including derivative works and methods based thereon as of the AIX Termination Date, June 13, 2003.

130. From and after the AIX Termination Date, any and all distributions of AIX by IBM is in violation of the Sublicensing Agreement.

131. To the extent that IBM continues to completely repudiate its obligations under the Sublicensing Agreement, plaintiff will sustain substantial continuing and ongoing damages. SCO is entitled to damages in an amount measured by the benefits conferred upon IBM by its ongoing, improper use of the Software Products. These damages include the full amount IBM receives as a result of its ongoing sales of AIX, including software, services and hardware.

132. Moreover, if IBM does not return or destroy all source and binary copies of the Software Products and/or continues to contribute some or all of these protected materials to open source, SCO will be irreparably harmed. As a result, SCO is entitled to a permanent injunction requiring IBM to return or destroy all source code and binary copies of the Software Products and/or prohibiting IBM from further contributions of the protected Software Products into open source.


(Breach of Sequent Software Agreement)

133. Plaintiff incorporates and re-alleges paragraphs No. 1-132, above.

134. As set forth above, SCO is the successor to AT&T under that certain Software Agreement originally executed by and between AT&T and Sequent designated as SOFT-000321. The Software Agreement specifies the terms and conditions for use of UNIX System V source code by Sequent.

135. With respect to the scope of rights granted for use of the System V source code under Section 2.01 of the Software Agreement, Sequent received the following:

[A] personal, nontransferable and nonexclusive right to use in the United States each Software Product identified in the one or more Supplements hereto, solely for Licensee's own internal business purposes and solely on or in conjunction with Designated CPUs for such Software Product. Such right to use includes the right to modify such Software Product and to prepare derivative works based on such Software product, provided the resulting materials are treated hereunder as part of the original Software Product. [Emphasis added.]

136. IBM has violated the grant of rights to Sequent under §2.01 of the Sequent Software Agreement by, inter alia, modifying and assisting others to modify the Software Products (including System V source code, derivative works and methods based thereon) for purposes other than Sequent and/or IBM's own internal business purposes. By actively supporting, assisting and promoting the transfer from UNIX to Linux, and using its access to UNIX technology to accomplish this objective, IBM is (a) using the Software Product for external business purposes, which include use for the benefit of Linus Torvalds, the general Linux community and IBM's Linux distribution partners, Red Hat, Inc. and SuSE Linux AG and its subsidiaries; and is (b) directly and indirectly preparing unauthorized derivative works based on the Software Product and unauthorized modifications thereto in violation of §2.01 of the Sequent Software Agreement.

Sequent designs and manufactures NUMA hardware. IBM purchased Sequent, and Sequent engineers helped port a new operating system to Sequent's hardware. SCO's products (UnixWare and OpenServer) do not run on Sequent hardware and are not sold for Sequent hardware. Why on earth does SCO think they are even involved here?

137. Sequent agreed in §2.05 of the Software Agreement to the following restrictions on use of the Software Product (including System V source code, derivative works and methods based thereon):

No right is granted by this Agreement for the use of Software Products directly for others, or for any use of Software Products by others.

138. IBM has breached Sequent's obligations under §2.05 of the Sequent Software Agreement by, inter alia, actively promoting and allowing use of the Software Products and development methods related thereto in an open and hostile attempt to destroy the entire economic value of the Software Products and plaintiff's rights to protect the proprietary nature of the Software Products. Particularly, IBM has caused all or materially all of DYNIX/ptx-based NUMA source code and methods, and RCU source code and methods, to be used for the benefit of Linux. But for the use by IBM of these protected UNIX methods in Linux development, the Linux 2.4.x kernel and 2.5.x kernel capacity to perform high-end enterprise computing functions would be severely limited.

Why is SCO claiming to own RCU technology?

Here's a little exchange from the Linux-kernel mailing list. In the first message, Linux kernel developer Andrea Arcangeli from SuSE rejects IBM's submission of RCU to Linux because the technology is covered by US Patent #5442758, as pointed out by Alan Cox of Red Hat.

The second message is from IBM employee Dipankar Sarma stating that IBM owns this patent, having purchased the inventor Sequent, and that IBM legal has reviewed it and approved its release under the GPL.

The third message is confirmation from Andrea Arcangeli that an IBM patent grant letter has been sent to both Linus and to him.

This exchange not only shows that IBM/Sequent has a patent on the RCU technology SCO is claiming ownership of, but is also an example of the Linux kernel maintainers performing due dilligence to the best of their ability to ensure the legal standing of the code submitted to them, even code submitted by IBM.

139. IBM has even gone so far as to publish the DYNIX/ptx copyright as part of the source code and documentation contribution of UNIX-derived RCU technology it has improperly made available to the open source community. The following copyright attribution is found in Linux kernel 2.4.x:

Copyright (c) International Business Machines Corp., 2001 This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. Author: Dipankar Sarma (Based on a Dynix/ptx implementation by Paul Mckenney.

It says it's based on a Dynix version implemented by a Sequent engineer. How does this involve SCO?

140. This publication of the RCU copyright is an example of IBM's blatant disregard of SCO's rights to control the use of the Software Product, including derivative works and modifications thereof, pursuant to §2.05 of the Sequent Software Agreement.

According to Bill Claybrook, one of the analysts SCO showed its case to:

DYNIX developed at Sequent years ago was derived from BSD 4.1 with patches from 4.2 and new code by Sequent. DYNIX/ptx, also developed at Sequent was really BSD code with System V wrappers. So the code was really still BSD code, the kernel code that is. It appears then that the NUMA kernel code was developed on the BSD code.

SCO appears to lack the basic understanding of copyright that any author must have in order to function. Sequent or IBM could add code to Unix and also add it to another product, and as long as it was original code they would be fully within their rights to do so. A copyright violation would be to put code SCO supplied into a non-Unix product. It would not be a copyright infringement for Sequent to use its own original code in multiple places, even if one of those places was a licensed Unix product and another was not.

Also, "based on" does not necessarily even imply a copyright violation. The parody "The Wind Done Gone" was based on the book "Gone With the Wind", but did not infringe upon its copyright. All Sequent is saying is that it referred to its own earlier work.

141. Sequent agreed in §7.10 of the Sequent Software Agreement to the following restrictions on transfer of the Software Product, including DYNIX/ptx as a derivative work of UNIX System V:

[N]othing in this Agreement grants to Licensee the right to sell, lease or otherwise transfer or dispose of a Software Product in whole or in part.

142. IBM has breached Sequent's obligations under §7.10 of the Sequent Software Agreement by, inter alia, transferring portions of the Software Product (including System V source code, derivative works and methods based thereon), including DYNIX/ptx source code, documentation and methods for NUMA, RCU and SMP technologies, to Linus Torvalds for open distribution to the general public under a software license that destroys the proprietary and confidential nature of the Software Products.

Sequent invented RCU, Sequent built NUMA hardware for almost two decades, and Caldera itself sponsored the early Linux SMP work during the 2.0 kernel series by donating hardware to Alan Cox.

143. Sequent agreed under §7.06(a) of the Sequent Software Agreement, to the following restrictions on confidentiality of the Software Product, including DYNIX/ptx as a derivative work of UNIX System V:

Licensee agrees that it shall hold all parts of the Software Products subject to this Agreement in confidence for AT&T. Licensee further agrees that it shall not make any disclosure of any or all of such Software Products (including methods or concepts utilized therein) to anyone, except to employees of Licensee to whom such disclosure is necessary to the use for which rights are granted hereunder. Licensee shall appropriately notify each employee to whom any such disclosure is made that such disclosure is made in confidence and shall be kept in confidence by such employee.

144. IBM has breached Sequent's obligation of confidentiality by contributing portions of the Software Product (including System V source code, derivative works and methods based thereon) to open source development of Linux and by using UNIX development methods in making modifications to Linux 2.4.x and 2.5.x, which are in material part, unauthorized derivative works of the Software Product, including but not limited to DYNIX/ptx-based NUMA technology, source code and methods, RCU source code and methods, and SMP source code and methods.

145. Export of UNIX technology is controlled by the United States government. Thus, SCO, Sequent, IBM and all other UNIX vendors are subject to strict export control regulations with respect to any UNIX-based customer distribution. To this end, Sequent agreed in §4.01 of the Software Agreement to restrictions on export of the Software Product (including System V source code, derivative works and methods based thereon), as follows:

Licensee agrees that it will not, without the prior written consent of AT&T, export, directly or indirectly, Software Products covered by this Agreement to any country outside of the United States.

This provision was later modified to allow export rights to several countries outside the United States. However, no permission has ever been granted by SCO or its predecessors to Sequent or IBM to allow either company to directly or indirectly make available all or portions of the Software Product to countries outside the United States that are subject to strict technology export control by the United States government: viz., Cuba, Iran, Syria, North Korea and Libya. IBM is ignoring and attempting to circumvent the export control restrictions that apply to UNIX as it accelerates development of Linux for enterprise use.

146. Thus, IBM has breached §4.01 of the Sequent Software Agreement by, inter alia, making extensive, advanced multiprocessor scaling functions of the Software Product, including NUMA technology, RCU technology, SMP technology and other derivative works and methods based thereon, available for free distribution to anyone in the world with a computer. As it relates to Linux 2.4.x and 2.5.x releases, IBM is indirectly making the Software Product and operating system modifications, particularly NUMA technology, RCU technology and SMP technology, available to countries and organizations in those countries for scaling single processor computers into multi-processor supercomputers that can be used for encryption, scientific research and weapons research.


(Unfair Competition)

147. Plaintiff incorporates and re-alleges paragraphs No. 1-146, above.

148. Plaintiff and its predecessors have built the UNIX System V Technology, the Unix Software Code, SCO OpenServer, UnixWare and their derivatives through very substantial efforts over a time span in excess of 20 years and expenditure of money in excess of $1 billion.

Similar things could be said about several of IBM's obsolete mainframe operating systems, DOS, OS/2, and the original MacOS. Moore's Law is entirely capable of depreciating a billion dollar investment down to a hundred thousand dollars in twenty years.[18]

SCO's most lucrative product today is is OpenServer, which SCO/Tarantella discontinued and SCO/Caldera initially didn't want to buy.

The article "Standards help Linux avoid Unix fate" talks about the fragmentation and infighting among vendors that slowly strangled the Unix market during the late 80's and early 90's, giving an opening to Microsoft's windows products. Novell's sale of Unix was a result of this fragmentation; by the time of the SCO/Tarantella purchase, it was already too late. (For more detail, see the new introductory material added to the second edition of the book "Life with Unix".)

149. IBM has engaged in a course of conduct that is intentionally and foreseeably calculated to undermine and/or destroy the economic value of UNIX anywhere and everywhere in the world, and to undermine and/or destroy plaintiff's rights to fully exploit and benefit from its ownership rights in and to UNIX System V Technology, the Unix Software Code, SCO OpenServer, UnixWare and their derivatives, and thereby seize the value of UNIX System V Technology, the Unix Software Code, SCO OpenServer, UnixWare and their derivatives directly for its own benefit and indirectly for the benefit of its Linux distribution partners.

IBM put out a competing product, which outcompeted SCO's obsolete trash in the marketplace. This is not illegal. SCO screwed up switching its core business from new technology (Linux) to old technology (OpenServer and UnixWare). And it did it entirely within the past three years. SCO can't blame IBM for its own stupid mistakes.

150. In furtherance of its scheme of unfair competition, IBM has engaged in the following conduct:

a) Misappropriation of source code, methods, trade secrets and confidential information of plaintiff;

b) Breach of contract;

c) Violation of confidentiality provisions running to the benefit of plaintiff;

d) Inducing and encouraging others to violate confidentiality provisions and to misappropriate trade secrets and confidential information of plaintiff;

e) Contribution of protected source code and methods for incorporation into one or more Linux software releases, intended for transfer of ownership to the general public;

f) Use of deceptive means and practices in dealing with plaintiff with respect to its software development efforts; and

g) Other methods of unlawful and/or unfair competition.

The only thing worth replying to here is (f). SCO refuses to reveal the specific source code it considers infringing, so that IBM and the Linux community can remove it. SCO has stated that it doesn't want the code removed, it wants a lawsuit. Yet SCO also claims to have made a good faith effort to resolve the issues. These are contradictory positions. Which company is using "deceptive means and practices"?

151. IBM's unfair competition has directly and/or proximately caused significant foreseeable and consequential harm to plaintiff in the following particulars:

a) Plaintiff's revenue stream from UNIX licenses for Intel-based processing platforms has decreased substantially;

Plaintiff's revenue stream has been decreasing for years. That's why the OpenServer and UnixWare product lines have passed through a chain of buyers. The year before the SCO acquisiton, Caldera bragged about their OpenLinux 2.3 being the fastest selling Linux product in Caldera history Abandoning Linux in favor of Unix was stupid, but it was their choice, even if it was one they assured their customers they would not do shortly before they did it.

b) As Intel-based processors have now become the processing platform of choice for a rapidly-increasing customer base of enterprise software users, plaintiff has been deprived of the opportunity to fairly exploit its market-leading position for UNIX on Intel-based processors, which revenue opportunity would have been very substantial on a recurring, annual basis but for IBM's unfairly competitive practices;

Caldera has never been a dominant Intel Unix vendor; it was a Linux company before acquiring the corpse of SCO's server division. SCO/Tarantella hasn't been a dominant Unix vendor for a decade, that's why it purchased UnixWare from Novell to prop up its product line.

The fall of the Unix clone Coherent in 1995 was detailed in the response to paragraph 39. SCO was one of several Unix vendors that the co-author of this document, Eric Raymond, tracked in his guide to Intel Unix clones before discontinuing it in 1994 because Linux had rendered everything else on the list obsolete and irrelevant. 1994 is also the year Caldera was founded, as a Linux vendor. (Ransom Love, who has been repeatedly quoted in this rebuttal, was a co-founder of Caldera.

c) Plaintiff stands at imminent risk of being deprived of its entire stream of all UNIX licensing revenue in the foreseeably near future;

Which is why SCO/Tarantella and Novell both sold the unprofitable white elephant that was the Unix source. It has nothing to do with IBM, and everything to do with the reason Tarantella discontinued OpenServer development, was looking into putting out its own Linux distribution, and was looking to partner with IBM in the first place. The ancient Unix technology it purchased is obsolete rubbish, useless on modern systems, has no hold over BSD, and Linux has been an independent line of development from its inception.

Leaving Linux out of it for a moment: before this year SCO/Caldera did not sell "System V". SCO/Tarantella also did not sell "System V", it sold UnixWare and OpenServer. They are not the same thing. Caldera's 2001 annual report, the first one issued after the acqusition of SCO's server division, does not mention System V, not even once.[19] System V was worthless by the end of the BSD lawsuit, which was settled shortly before Novell rid itself of Unix. Everybody who wanted System V licenses had already purchased "irrevocable, perpetual" versions long before Caldera got ahold of it. The few old-time licensees who had not bought out their royalty obligations were paying royalties to Novell, not to SCO.

Neither UnixWare nor OpenServer were initiated by then-owners of AT&T's System V rights. They were seperate lines of development, initiated by normal Unix Licensees, and not tied to AT&T's System V rights in any special way. Novell had not yet purchased USL when it started UnixWare, and SCO and Microsoft were normal Unix licensees (not owners) when Xenix was renamed to OpenServer. UnixWare and OpenServer were always minor versions of Unix, the versions belonging to other vendors have always been more important and more lucrative: Sun's Solaris, HP's HP-UX, and IBM's AIX being three surviving profitable examples. All of these companies have purchased "irrevocable, perpetual" licenses to AT&T's old Unix codebase, and will never owe SCO another dime for it.

The proprietary Unix industry is still worth billions of dollars annually, money made off of Solaris, HP-UX, AIX, and other proprietary Unix variants. SCO's lack of participation in this industry is entirely SCO's own fault: its products are inadequate, its marketing strategy is completely ineffective, its ability to negotiate business deals is nonexistent, and its poor financial health and history of declining revenue have discouraged large corporate customers from partnering with a company that may not be around in future to support its products. This is why SCO/Tarantella wanted to sell its Unix business. Caldera was stupid to shift focus from its Linux business to its Unix business, but it has only itself to blame.

UnixWare is trailing edge technology, approximately on a par with the Linux 2.2 kernel from 1999. OpenServer is so obsolete SCO/Tarantella attempted to discontinue its development in 1999, but it has an entrenched userbase of point of sale systems whose owners don't care how obsolete the technology is, and are so resistant to change they're willing to pay a premium to avoid having to upgrade. System V is a museum piece, an obsolete relic primarily of interest to historians.

SCO's business strategy has been to milk its installed base, surviving off a steadily declining revenue stream and attempting to cut costs to achieve profitability. They must have known there was no future in that strategy when they set off down that path, and it is dishonest of them to try to blame others for their own mistakes.

d) Plaintiff has been deprived of the effective ability to market and sell its new UNIX-related improvements, including a 32-bit version of UNIX for Intel processors developed prior to Project Monterey, a 64-bit version of UNIX for Intel processors based on Project Monterey, and its new web-based UNIX-related products, including UNIX System VI;

Notice that they don't say they were deprived of the ability to market and sell their products, just of the effective ability to do so. This is the language of incompetents, trying to blame the fruits of their incompetence on IBM.

The main feature of their new 32-bit version of Unix for Intel processors is the "Linux Kernel Personality", and why would someone buy UnixWare to run Linux programs when they can simply use Linux instead? There is no market for Monterey systems, only 3500 Itanium systems shipped in the whole of 2002; this is not IBM's fault.

I haven't been able to find out anything about "UNIX System VI". The top hit in Google is a decade old humor piece. It doesn't help that "VI" is the name of the BSD text editor command, making the name of SCO's product almost impossible to search for. This horrible marketing decision simply cannot be blamed on IBM.

e) Plaintiff has been deprived of the effective revenue licensing opportunity to transfer its existing UNIX System V customer base to UNIX System VI; and

Lack of customer demand for a new product is not proof of a conspiracy. The last new product Caldera released for which there was significant customer demand was OpenLinux 2.3, yet they ignored this to focus on obsolete versions of Unix.

f) Plaintiff has been deprived of the effective ability to otherwise fully and fairly exploit UNIX's market-leading position in enterprise software market, which deprivation is highly significant given the inability of Microsoft Windows NT to properly support large-scale enterprise applications.

The Unix market leader is Solaris. How did SCO expect to exploit Solaris's position as Unix market leader, when Sun has bought one of the "irrevocable, perpetual" licenses to Unix and will never owe SCO another nickel for it?

152. As a result of IBM's unfair competition and the marketplace injury sustained by plaintiff as set forth above, plaintiff has suffered damages in an amount to be proven at trial, but no less than $1 billion, together with additional damages through and after the time of trial foreseeably and consequentially resulting from IBM's unfair competition in an amount to be proven at the time of trial.

153. IBM's unfairly competitive conduct was also intentionally and maliciously designed to destroy plaintiff's business livelihood and all opportunities of plaintiff to derive value from its UNIX-based assets in the marketplace. As such, IBM's wrongful acts and course of conduct has created a profoundly adverse effect on UNIX business worldwide. As such, this Court should impose an award of punitive damages against IBM in an amount to be proven and supported at trial.

Caldera was a Linux vendor. At the time of trial, SCO/Caldera sold the product SCO OpenLinux. SCO has never disputed that IBM's actions have helped Linux, presumably including SCO's OpenLinux product. Up until about a year ago, OpenLinux was at least nominally SCO's main product. How can improving SCO's main product be considered a malicous conspiracy against SCO? Even under SCO's own theory of code misappropriation, the harm to SCO came from SCO's poor business decisions. Other Unix vendors are making billions of dollars, and other Linux vendors are making billions of dollars. SCO's incompetence is SCO's fault.


(Interference with Contract)

154. Plaintiff incorporates and re-alleges by reference paragraphs 1-153, above.

155. SCO has contracts with customers around the world for licensing of SCO OpenServer and UnixWare.

SCO has been dismantling its reseller channel for some time now as part of its cost cutting initiative. This has unsurprisingly impacted the gross revenue it receives from that reseller channel. The appropriate colloquial expression is "duh".

156. IBM knew and should have known of these corporate software licensing agreements between SCO and its customers, including the fact that such agreements contain confidentiality provisions and provisions limiting use of the licensed object-based code.

157. IBM, directly and through its Linux distribution partners, has intentionally and without justification induced SCO's customers and licensees to breach their corporate licensing agreements, including but not limited to, inducing the customers to reverse engineer, decompile, translate, create derivative works, modify or otherwise use the UNIX software in ways in violation of the license agreements. These customers include Sherwin Williams, Papa John's Pizza, and Auto Zone, among others.

Reverse engineering is legal. And is SCO really saying that Papa John's Pizza would breach license obligations if IBM told it to?

158. IBM's tortious interference has directly and/or proximately caused significant foreseeable damages to SCO, including a substantial loss of revenues.

SCO/Caldera shifted focus from the fastest selling product in its history to a product line with a history of declining revenues.

159. IBM's tortious conduct was also intentionally and maliciously designed to destroy plaintiff's business livelihood and all opportunities of plaintiff to derive value from its UNIX-based assets in the marketplace. As such, this Court should impose an award of punitive damages against IBM in an amount to be proven and supported at trial.

It's ironic that SCO has allowed its attack on Linux and IBM to be funded by Microsoft, a company which is explicitly out to destroy plaintiff's business livelihood by putting "Windows everywhere". Then again, barring antitrust concerns about a convicted predatory monopolist, this is not illegal. Companies out-competing each other in the marketplace is a fundamental part of our capitalist economy. That said, IBM didn't try to destroy SCO, it parnered with SCO/Tarantella to develop Monterey, and resold SCO/Caldera's Linux to IBM customers. But even if it had, that would not be actionable in and of itself.


(Misappropriation of Trade Secrets—Utah Code Ann. §13-24-1 et seq.)

160. Plaintiff incorporates and re-alleges paragraphs No. 1-159, above.

161. Plaintiff is the owner of unique know how, concepts, ideas, methodologies, standards, specifications, programming, techniques, UNIX Software Code, object code, architecture, design and schematics that allow UNIX to operate with unmatched extensibility, scalability, reliability and security (hereinafter defined as “SCO's Trade Secrets”). SCO's Trade Secrets provide SCO with an advantage over its competitors.

Nothing SCO has is "unique". The license the Unix trademark from somebody else, they pass on Unix royalties to Novell, many companies have an "irrevocable and perpetual" license to System V, and anything System V does BSD can do without encumbrance. SCO's new version derived from System V is one of dozens of derived versions, all extended far beyond the original 20 year old System V code.

SCO has no patents, they don't own the trademark, copyright won't serve them and the only contract they have with the Linux community is the General Public License, which SCO is the one violating. So they fall back on trade secrets, which aren't secret anymore after 30,000 different source licenses were issued over a period of 30 years, many of them even explicitly licensed for educational purposes. This is one area where anecdotal evidence DOES help: a trade secret only has to be leaked once in order to no longer be secret. Eric is collecting these stories on his no secrets page. If you have a story to contribute, contact Eric.

Even without the 30 years and 30,000 licensees, SCO's trade secret claims are weak for other reasons. The judge in the BSD lawsuit refused to issue a preliminary injunction because he didn't see any trade secrets at risk. SCO/Caldera itself put Unix System III and earlier up for download on its web page. And of course SCO shipped Linux 2.4.14 (and more recent versions as part of UnitedLinux), and profited from doing so, apparently without performing any sort of due dilligence upon what exactly it was shipping. Now they claim it's a secret.

The real reason SCO is focusing on trade secrets is that it gives them an excuse (however weak) to pretend to have evidence without having to actually produce any in public. This plays well in the press, but is unlikely to help them in court. The Open Source community has an old saying, "Shut up and show me the code".

SCO of course has not indicated any specific trade screts it considers to have been misappropriated. Presumably SCO will be forced to reveal in court precisely what trade secrets it thinks IBM misappropriated, and expert witnesses will be allowed to research the matter and individualy prove that each one was already widely known in the industry and academia. Considering that SCO is claiming ownership of RCU, which Sequent/IBM invented and owns a patent on, this is unlikely to be too difficult.

162. SCO's Trade Secrets are embodied within SCO's proprietary SCO OpenServer and its related shared libraries and SCO's UnixWare and its related shared libraries.

163. SCO and its predecessors in interest have expended over one billion dollars to develop SCO's Trade Secrets.

Depreciated according to Moore's Law. Digital Equipment Corporation spent more than a billion dollars developing the VAX, which is no longer being sold.

Caldera purchased SCO's Unix business for $36 million. This included the OpenServer product line developed by SCO/Tarantella and Microsoft, the UnixWare product line developed by Novell, and the ancient versions of Unix (System V and earlier) developed by AT&T. However much this code initially cost to develop, Caldera put System III and earlier up as a free download on their website, and OpenServer was at the time a discontinued product line for which the development of new features had been ruled out by SCO/Tarantella, one which Caldera initially did not want to purchase from SCO. And Caldera's publicly stated interest in the acquisition of SCO was to get a distribution channel through which to sell Caldera OpenLinux. Therefore, only a tiny fraction of the $36 million could be attributed to System V. In its last annual report, Caldera itself put the value of all the Unix technology it acquired from SCO (OpenServer, UnixWare, System V, System III, etc...) at around $1 million.

Depreciation aside, it's also interesting to note that the $1 billion dollar figure keeps getting re-used. SCO thinks it suffered a billion dollars damages from copyrights, and a billion dollars damages from trade secrets. Except this is the same billion dollars approached from two different directions; it didn't get spent twice. SCO has also repeatedly made and withdrawn allegations about patents in the trade press. Presumably, patents would be a third re-use of the same billion dollar figure, if Novell hadn't retained the Unix patents back when it sold its Unix contracts to SCO and gave the Unix trademark to The Open Group.

164. IBM, through improper means acquired and misappropriated SCO's Trade Secrets for its own use and benefit, for use in competition with SCO and in an effort to destroy SCO.

SCO is destroying itself quite nicely without any help from IBM. The first profitable quarter SCO/Caldera ever had as a publicly traded company was due to Microsoft paying them to support the suit against IBM.

165. At the time that IBM acquired access to SCO's Trade Secrets, IBM knew that it had a duty to maintain the secrecy of SCO's Trade Secrets or limit their use.

At which time? Is it talking about the System V licensing back in the 1980's or about the Monterey project? Be specific. And what trade secrets does SCO claim IBM leaked? At SCOForum 2003, SCO showed the audience A) a piece of code submitted by SGI and quickly ripped out again because it was bad code and unnecessary (long before SCOForum), B) a section of the Berkeley Packet Filter, from BSD. That's the best they could do at their own annual developer conference.

166. SCO's Trade Secrets derive independent economic value, are not generally known to third persons, are not readily ascertainable by proper means by other persons who can obtain economic value from their disclosure and use, and are subject to reasonable efforts by SCO and its predecessors to maintain secrecy.

SCO must not only exert reasonable efforts, it must have a perfect record. A trade secret that legitimately leaks even once is no longer a secret. If SCO or its authorized agents screwed up even once, so that the recipient of the code was given it by someone who was authorized to have it, but the recipient was not bound by a non-disclosure agreement, then the secret is no longer a secret.

167. The acts and conduct of IBM in misappropriating and encouraging, inducing and causing others to commit material misappropriation of SCO's Trade Secrets are the direct and proximate cause of a near-complete devaluation and destruction of the market value of SCO OpenServer and SCO UnixWare that would not have otherwise occurred but for the conduct of IBM.

SCO is completely wrong that the devaluation of OpenServer and UnixWare would not have occured without IBM. Consolidation in the Unix business was already occuring when the Mark Williams Company, makers of Coherent, went out of business in 1995. The consolidation of System V, UnixWare, and OpenServer under one roof is an example of what the industry refers to as "dinosaurs mating", and that consolidation happened in 1995, long before IBM's involvement with Linux.

168. Pursuant to Utah Code Ann. §13-24-4, plaintiff is entitled to an award of damages against IBM in the following amounts:

a) Actual damages as a result of the theft of trade secrets; together with

b) Profits from IBM's Linux-related business on account of its misappropriation through the time of trial; together with

c) Additional foreseeable profits for future years from IBM's Linux-related business on account of its misappropriation in an amount to be proven at the time of trial.

169. Because IBM's misappropriation was willful, malicious, and in reckless disregard of plaintiff's rights, SCO is entitled to an award of exemplary damages against IBM in an amount equal to two times the amount of damages, pursuant to Utah Code Ann. §13-24-4(2).

On the evidence, IBM did not take any willful, malicious, or reckless actions towards SCO. IBM wasn't paying the slightest bit of attention to SCO. When SCO's stock was trading at $1/share, IBM could have purchased the entire company for less money than it cost to stage the annual Comdex expo in 1999, another business enterprise that has been hit by hard times. Is Comdex expected to sue IBM next, just because IBM is still making money and they aren't?

170. Plaintiff is also entitled to an award of attorneys' fees and costs in an amount to be proven at the time of trial pursuant to Utah Code Ann. §13-24-5.

And if IBM wins, it should get attorneys' fees from SCO for this harassment.

Prayer for Relief

WHEREFORE, having fully set forth its complaint, plaintiff prays for relief from this Court as follows:

1. For damages under the First Cause of Action for breach of the IBM Software Agreement in an amount not less than $1 billion, together with additional damages through and after the time of trial foreseeably and consequentially resulting from IBM's breach, in an amount to be proven at the time of trial;

There's that magically multiplying billion dollars again. SCO bought a distribution channel for $36 million which had some old Unix technology sticking to it, and after scraping that off it's trying to cash it in for $1 billion like it was a lottery ticket.

SCO/Caldera's gross revenue off of the purchased Unix products, while consistently declining every year, has totalled more than $36 million. The fact their net revenue on it has been negative is their own fault, in gross terms it paid for itself. (Barely, and mostly due to the fact they changed their mind about purchasing OpenServer.)

If SCO did get an award from IBM, would this bar it from going and suing other Linux companies? This does seem to be its new business model, it's moved out of the technology industry and into the spurious lawsuit industry. If it gets an award from IBM (say, one dollar, a pat on the head, and cab fare home), would it be barred from suing anybody else over the same code? Or could it keep on suing anyone and everyone who uses Linux, a product that has clearly outcompeted SCO's own products in the marketplace?

2. For a permanent injunction under the First and Second Causes of Action requiring IBM to return or destroy all source code and binary copies of the Software Products and/or prohibiting IBM from further contributions of the protected Software Products into open source;

IBM's AIX business makes more money than the whole of SCO. IBM is in the process of slowly migrating its own AIX customers to Linux, as is most of the rest of the industry. The fact SCO/Caldera has gone in the opposite direction from the rest of the industry is due to Caldera's stupidity.

Law works on precendent, and the clear precedent here is the BSD case. A settlement was made whereby any specific and significant code snippets the plaintiff can prove to be in violation of its copyrights (if any) were removed from future versions of the other project, then the two code bases were cleanly separated and declared unencumbered. The precedent here is almost inescapable; the BSD case was about the System V code base, and the plaintiff there was SCO's predecessors in interest AT&T, USL, and Novell.

3. For restitution under the First and Second Causes of Action in an amount measured by the benefits conferred upon IBM by its ongoing, improper use of the Software Products, including the full amount IBM receives as a result of its ongoing sales of AIX, including software, services and hardware;

If SCO/Caldera asks for IBM's AIX revenue, than IBM should ask SCO/Caldera to give back every dime of Linux revenue it made from the sale of versions of Linux that contain any IBM code. (Either SCO accepts the Gnu Public License under which it shipped Linux up through and including version 2.4.14, or it forfeits all that Linux revenue to IBM and the rest of the Linux development community. IBM is, after all, spending $1 billion a year on Linux, and Caldera profited from Linux as its core business for many years.)

4. For damages under the Third Cause of Action for breach of the Sequent Software Agreement in an amount not less than $1 billion, together with additional damages through and after the time of trial foreseeably and consequentially resulting from IBM's breach, in an amount to be proven at the time of trial;

There's that multiplying $1 billion again. It wants every dime AT&T ever spent on Unix back in the 1980's paid to it by IBM, and then it wants another equal payment from Sequent. SCO is explicitly asking for more money in damages than it says was spent on the entire history of Unix development.

5. For relief under the Fourth Cause of Action in an amount not less than $1 billion, for unfair competition arising from common law, and damages for violations thereof, together with additional damages through and after the time of trial;

The magic multiplying billion resurfaces. SCO seems to believe that all competition is "unfair", and that its rightful place as king of the world was taken from it via a conspiracy, and the company to blame is obviously the company that invented the IBM PC that it's been selling Unix for all these years.

6. For relief under the Fifth Cause of Action in an amount to be proven at trial for tortious interference, together with additional damages through and after the time of trial;

Back to the claim "IBM turned Papa John's against us". Presumably, some executives from the named companies will be willing to testify on IBM's behalf.

7. For relief under the Sixth Cause of Action in an amount to be proven at trial for misappropriation of trade secrets arising from Utah Code Ann. §13-24-1 et seq., together with additional damages through and after the time of trial;

The judge in the BSD trial didn't see any AT&T System V trade secrets present in the BSD code base, it's unlikely that 20 years later AT&T System V trade secrets would be found in Linux, especially after Caldera put System III up for download on its website and sold Linux 2.4.18 with source code under the GPL.

8. For a permanent injunction under the Third Cause of Action to prohibit IBM from further contributions of the protected Software Products into open source;

This essentially repeats entry number 2.

9. For punitive damages under the Fourth and Fifth Causes of Action for IBM's malicious and willful conduct, in an amount to be proven at trial;

IBM took action to improve a product (Linux) which was Caldera's core business. SCO/Caldera alleges they took code from a product about which SCO/Caldera's CEO Ransom Love had promised "We won't make Unix proprietary." Elsewhere, responding to the question "What will happen to OpenServer?", SCO CEO Ransom Love said,

"We have more than two million installations of OpenServer. The operating system is in maintenance mode now, so there will be no more major enhancements. But what we plan to do is to take the OpenServer technology to Linux, probably with some sort of open source source licence..."

Under those circumstances, how could IBM possibly have acted maliciously? Confused by mixed signals, certainly, but not malicious intent.

10. For exemplary damages pursuant to Utah Code Ann. § 13-24-1 under the Sixth Cause of Action in an amount equal to twice the award under the Sixth Cause of Action for misappropriation of trade secrets;

11. For attorneys' fees and costs as provided by Utah Code Ann. § 13-24-5 and/or by contract in an amount to be proven at trial; and

12. For attorneys' fees, costs, pre- and post-judgment interest, and all other legal and equitable relief deemed just and proper by this Court.

Jury Trial Demand

SCO demands trial by jury on all issues so triable.

      DATED this ___ day of June, 2003.

                                    HATCH, JAMES & DODGE
                                    Brent O. Hatch
                                    Mark F. James

                                    BOIES, SCHILLER & FLEXNER
                                    David Boies
                                    Stephen N. Zack
                                    Mark J. Heise

                                    Attorneys for The SCO Group, Inc.

Plaintiff's address:

355 South 520 West
Lindon, Utah 84042

The complaint ends here. The following text is editors' appendices.

A. The Decline and Fall of Caldera

So, how did it come to this? How did Caldera, a senior Linux vendor whose core business was developing, selling, and supporting a Linux distribution ever since its founding in 1994 (the same year as leading Linux vendor Red Hat), turn into a near-bankrupt proprietary Unix company whose main product is spurious lawsuits about 20-year old technology?

The change happened gradually. When Caldera purchased the server division of SCO in 2000, they were primarily buying a reseller channel through which to sell their Linux products (Caldera's core business at the time). [20] SCO's Unix technology came along for the ride because that's what the distribution channel was currently selling, and switching everything over to Linux would take time.

As Caldera's CEO Ransom Love explained in an August 17, 2001 interview in response to the question, "How heavily will you rely on resellers?":

We have a large OEM team directed towards OEMs and resellers, and we see the channel as critical to business adoption of Linux. The reseller is the one who takes our products, who understands them and who turns them into solutions for business customers. We now need VARs more than ever, even though we make Linux as easy to install and manage as we can.

For the business customer buying open source software, I don't believe in the direct market, I believe in the channel. Open source software can be overwhelming to these customers.

Right before the acquisition, Caldera released OpenLinux 2.3, which their October 4, 1999 press release touted as the Fastest Selling Linux Product in Caldera History. They were doing just fine with Linux, and had no reason to switch focus away from their core business as a Linux distributor. So why did they?

Caldera's switch from a Linux vendor to a Unix vendor occurred because the distribution channel it had acquired from Tarantella was more than it could control. For Caldera the experience was like a child wrestling with a fire hose. The entrenched customer and reseller relationships were many years old and completely inflexible, and although they provided more gross revenue than the rest of Caldera combined, they lost money in net terms because they cost so much to operate. Caldera tried to cut expenses to pare the channel down to something they could manage, and that would actually be profitable, but the gross revenue declined as the expenses declined, and profitability remained tantalizingly out of reach. Eventually, some of them came to understood why Tarantella had been so happy to sell their server business so cheaply.

Caldera's management changed as well, Caldera CEO and co-founder Ransom Love brought in Darl McBride, who soon replaced Love as CEO. Renaming Caldera to SCO was one of McBride's first initiatives as CEO, as explained in an interview DesktopLinux conducted with McBride at SCO/Caldera's 16th annual Forum:

DL: And when did the idea of changing the name back to SCO come up?

McBride: It came up early on. The very first week I was there...

I had three books of press clippings over the past few months. Ninety-nine percent of the clippings related to Linux. Cool. Now, through the fifteen discussions, the identity crisis related to the fact that everyone talked about our core business. Ninety percent of our business comes from UNIX. And when you get out there, you have all these SCO re-sellers who don't attach to the corporate message.

It was clear we have an identity crisis we've got to fix. And there are multiple ways we can fix it. We could pound on the SCO re-sellers until they "get it." Obviously, that doesn't make sense, because they are generating good revenue. So you want to promote them, encourage them to go out selling, so that when Linux is ready for Prime Time, which it is really not right now, you are able to drop into that channel and get traction with the Linux products.

(In the same interview, Darl McBride also explains his history with Linux and open-source software, and how Ransom Love recruited him to come to Caldera.)

Along the way, Caldera forgot about repurposing the channel to sell Linux. Ignored, their Linux product atrophied to a small fraction of their revenues. Caught up in the ongoing cost-cutting, Linux development at Caldera stagnated, and what was left was outsourced to another Linux company (SuSE) through a partnership called UnitedLinux (of which SCO/Caldera still owns 25%[21]).

In a wide-ranging August 15, 2002 interview with ZDNet, McBride looks back on how SCO/Caldera had backed away from Linux under his stewardship:

ZDNET: Just a few years ago Caldera's Linux distributions were a major part of its product lineup. These days Caldera's main push is Unix. What happened to Linux?

McBride: The problem was, they were spending four marketing dollars to get one revenue dollar. No matter how much you raise on your IPO, you can't do that forever. I think acquiring SCO was a really smart thing. There was a view that the SCO business was going to go away, but actually it's been the opposite. The constricting of the economy has led to the effect that companies are not immediately swapping out core platform products, especially when they work as reliably as an SCO product does. We just signed a big licensing deal with McDonald's last week, BMW--these guys are coming back in droves right now, signing up to continue on with the platform. A year ago the strategy was "Let's move all of our SCO customers onto Linux." We have a much different view of the world this year. We have an operating systems group here, and we have OpenServer, UnixWare, and UnitedLinux, and we're going to let customers choose whatever they want. Each OS is tuned for a different market--UnixWare is great for telcos, for example, and we think UnitedLinux is going to be better from a back-office point of view. At the point of sale with a lot of apps surrounding it, that's where SCO is doing well. What we want to do on top of that is add a set of open-source applications that wrap around that installed base, and try to make money on doing some of the higher-level services and apps as well. In a nutshell, that's where we're trying to go with the company.

(The Ziff-Davis interview is primarily about SCO/Caldera's role in the formation of UnitedLinux, and its plans for Linux going forward, but McBride also opens with comments about his study of Caldera's ancient IP portfolio, and how ignorant he was about Unix history when he took office.)

Under McBride's leadership, Caldera gradually lost its corporate identity entirely and became the shriveled husk of an unprofitable Unix business that Tarantella, Novell, and AT&T had sequentially disposed of. You might think that the remains of Caldera then had an enormous sour grapes reaction to Linux, decided that if it couldn't profit from it anymore nobody else should either, and started suing people. But the reality was, they simply ran out of money and repeated the last thing they'd done that produced a big pile of cash: suing a big company over a long-obsolete product they'd purchased from the corpse of a failed business.

Caldera had never been profitable since their IPO. For years, they had been coasting on the proceeds from the settlement Microsoft paid them in their suit over Digital Research's DR-DOS. In 2000, shortly after Microsoft's antitrust conviction as an abusive monopolist, Microsoft settled a lawsuit Caldera had filed against it, which claimed that Microsoft put DR-DOS out of business by unfairly leveraging Microsoft's monopoly. After losing to the DOJ, Microsoft was willing to give Caldera money to just go away and leave it alone.

The amount of the settlement was secret, but the answer (a little over $150 million) is fairly easy to work out from the appropriate SEC filings. [22] (Interestingly, however Caldera did not get this money. Due to a corporate shell game detailed an an article in The Register, the Microsoft settlement was swallowed by Caldera's main shareholder, Ray Noorda's Canopy group. [23])

Meanwhile, Caldera got $30 million of venture capital financing from Sun, SCO, Citrix, and Novell, and went on to hold an IPO that netted them $65.1 million. (This is an impressive amount of money, but the venture capital and IPO combined were significantly less than the legal settlement from Microsoft.)

This is the money they used to purchase SCO's server division later in 2000 (plus an $18 million loan from Canopy), and also the money that funded their operations for the next three years. But at the start of 2003 the money ran out, and since they couldn't hold another IPO (certainly not with their stock trading at around $1/share, which it was at the time), they looked around for another big company with deep pockets to blame for the failure of somebody else's antiquated technology, which they had since purchased. They forgot that last time, their lawsuit had an important element that this one lacks: a case.

Of course SCO knows that since they go through cash like toilet paper, a lump-sum payment merely buys them time. What they're really looking for (or what puppeteer Canopy is looking for) is another buy-out, just as Tarantella, Novell, and AT&T before them handed off the unprofitable System V hot potato, under the "bigger sucker" theory.

SCO/Caldera's transformation from a Linux vendor to a Unix vendor is very recent. In February 2001, a letter outlining Caldera's official position from Nancy Pomeroy, Caldera's Director of Corporate Communications, was written to reassure the Linux community:

Caldera Systems assures the Linux community that we are not pulling out of Linux - more specifically that our OpenLinux workstation and server products will continue. What will change is where we focus our sales efforts. Caldera's focus has always been "Linux for business." With the pending acquisition of the Professional Services and Server Software divisions of The Santa Cruz Operation (SCO) we will unify UNIX and Linux for business. This business focus means selling our products through the channel and OEMs. In fact, it's an important reason why we view the SCO acquisition as strategic. SCO has a strong presence in the channel. When the acquisition is complete we will have a combined Linux and UNIX channel of over 22,000 resellers.

At this time we are expending significant resources developing OpenLinux workstation and server solutions that incorporate the 2.4 kernel. Given our business focus, the 2.4 kernel integrations in our workstation and server solutions will continue to demonstrate our commitment to stability, ease-of-use and manageability.

More succinctly, Caldera's CEO Ransom Love promised after the SCO acquisition:

Ransome Love, chief executive at Caldera, said: "What the world needs now is a company that can give Linux a global infrastructure and a commitment to open source computing."

"We won't make Unix proprietary," said Love. "The future is an open internet platform."

SCO/Caldera still has a web page devoted to its contributions to the open-source community. It's a fitting memorial to the company they once were, and are no longer.

B. Sizing the Unix Market

If IBM is, as SCO alleges, engaged in a malign conspiracy to destroy Unix, it is a remarkably ineffective one. Other companies with more competent management than SCO's are making quite a lot of money selling Unix. IBM itself does a land-office business in AIX, and it is no better than third in the market.

A recent article indicates that HP made more money from Linux last year as IBM did, and goes on to say that the Unix market (excluding Linux sales) is still worth over $4 billion annually, and that Sun and HP are tied for first place with IBM coming in third. The fact that SCO's total gross annual Unix product revenue was only $53 million in 2002 (see page 35) is not due to lack of opportunity in the multi-billion dollar Unix market.

The fact that the Linux business is flourishing has yet to seriously affect Sun, HP, or IBM's high end “enterprise” Unix business. Most industry observers expect commodity Linux to climb into that market eventually, but in a relatively gentle transition that most old-line Unix companies seem to be handling gracefully.

SGI (Silicon Graphics, Inc.) provides an instructive example of this transition. It introduced x86-based workstations at the end of the 1990's, but instead of porting their existing Unix product (Irix) to the Intel platform they chose to use Linux instead. SGI remains an ardent supporter of open source software development, and Linux in particular.

The intense competition SCO has experienced from Linux may more accurately be put down to the fact that SCO is not in the "enterprise" Unix business, and never has been. It is a low-end PC Unix vendor, primarily selling cash registers to fast food and retail outlets, which is a task for which a modern Palm Pilot would be overkill. Linux needed no help to flourish on low-end PCs for most of a decade before IBM even noticed its existence, and SCO had plenty of competition besides Linux in the low-end PC Unix market back in 1994. SCO/Caldera may finally have figured out why SCO/Tarantella wanted to get out of the low-end PC Unix business. The market has commoditized and margins are razor-thin there; neither of these things are IBM's fault.

C. Moore's Law

SCO doesn't bother to explain Moore's Law, which is important for understanding both why there was no special insight in SCO/Tarantella's business plan and why the System V technology SCO touts so highly is nearly valueless today.

Moore's Law was originally an observation about the economics of integrated circuit manufacturing technology (published on April 19, 1965, some years before Gordon Moore [24] became a co-founder of Intel) but has since been generalized into its modern form: that the amount of computing power available at a given price point doubles every eighteen months. (This also means that a given amount of computing power — memory, processor speed, disk drive space, etc — will generally be half as expensive a year and half later.)

Moore's Law was almost 15 years old by the time of SCO's founding; widely known and providing SCO with no unique insights. Gordon Moore famously updated his law (to be more accurate based on new information) for its tenth anniversary, at the IEEE International Electron Devices meeting in 1975. Here is a link to a 1979 commentary on Moore's Law from Stanford University's Graduate School of Business.

Moore's law is not limited to Intel chips, it applies equally to competing processors like RISC systems or Motorola's 68k processors. Motorola's 68k processor line was the main competitor to Intel's x86 line from the time of the IBM PC's introduction until Motorola, IBM, and Apple's joint development of the Power PC processor line in the early 1990's. Although IBM chose Intel's 16-bit x86 processors for the IBM PC in 1981, most contemporary Unix vendors used the more powerful 32-bit 68k processors for their workstations. (Many other contemporary home computer vendors also chose Motorola's 68k for their home computer products: Apple used a 68k for the Macintosh, Commodore used a 68k for the Amiga, Atari used a 68k for the ST... Very few systems but IBM's PC and its clones used Intel's x86 line at the start of the 1980's.) The IBM PC took off not because of superior technology, but because of IBM's brand name.

Even Intel didn't initially see IBM's PC as particularly interesting, it was simply trying to win business away from Motorola's 68k. In an interview with PC magazine, Intel's Gordon Moore (of Moore's Law) explained:

That was a period of time when we were competing very strongly with Motorola for design wins. They had a good part out on the market, and we were concerned that they were going to run away with a whole generation of designs, so we put together a very aggressive marketing program, where we wanted to get 2,000 design wins. I think it was in the '79-'80 time period, over something like a year span, that time, and went aggressively out to pitch every design we could. Well one of those designs happened to be the IBM PC. In fact the salesman that was calling on the account wasn't even told what the project was... eventually we found out it was the PC. But even IBM expected only to sell a few hundred thousand of them over the lifetime of the product. And while we thought it was a significant design win, it was one of something like 3,000 that we had in that time. And I certainly didn't realize that it was going to be the future of Intel, even when the IBM PC got announced.

Moore's Law provided Intel with no guarantee of success in the marketplace against other microprocessor competitors. In the late 1970's, the company Zilog, founded by ex-Intel employees, out-competed Intel in the 8-bit processor space before Intel moved upmarket to 16-bit and 32-bit processors. (This is similar to the modern relationship between Intel and Advanced Micro Devices (AMD), maker of the x86 compatible "Athlon" processor line.)

Another interesting point is that Moore's Law can be used to calculate depreciation. The price of existing computer technology is cut in half every 18 months, therefore ten year old technology has lost half its price over six times, turning a million dollar hardware investment into about ten thousand dollars.

The changes in hardware demand corresponding changes in software. Networking once meant dial-up modems, now it means gigabit ethernet. An impressive graphics display was once 640 pixels wide and 480 pixels tall. Now even laptop displays can have 1600 by 1200 resolution (requiring 64 megabytes of memory just for the video display, more than the entire system would have had five years earlier). Managing vastly different amounts of storage, memory, and processing power demands different algorithms, leading to an often overlooked result: Software also depreciates following Moore's Law.

The effects of Moore's Law on software can be seen by the fact that new versions of the same software come out every few years, and that even spectacularly popular applications like Lotus 1-2-3 and WordPerfect can stop selling a few years after their release unless successfully updated. Version numbers like Word Perfect 8, Red Hat 9, and MacOS X (roman numeral 10) are common because new versions must be released every year or two to remain competitive. Industry insiders refer to this as the upgrade treadmill.

The single most profitable line of software in the world is Microsoft's dominant operating systems (mentioned in paragraph 19 of this complaint). According to Microsoft's License Availability Roadmap, the company's general policy is to stop selling windows versions after 5 years. MS-DOS (the dominant operating system through the 1980's), Windows 3.1, and Windows 95 have already been retired. Windows 98 is no longer available retail, and will no longer be available to computer manufacturers after November, 2003. Windows ME, windows 2000, and even the current Windows XP all have announced retirement dates. Microsoft spends billions of dollars developing new versions of windows, yet discards each version as worthless after only five years. (Microsoft even switched from version numbers to expiration dates, using the year of release in products like "Windows 95" and "Office 97" to let people more easily know when it was time to upgrade.)

For a broader example, one of the most profitable software niches is the multi-billion dollar video game industry, which makes more money than Hollywood. According to this article on The Business of Computer Games

What is the shelf life of a typical high-budget retail game? Not long. Somewhere between two weeks and six months... A typical game will make most of its sales during the first quarter, a few more sales during the second quarter, and end up in the bargain bin in the third quarter. Expect to get one good royalty check, if you are lucky, and if your publisher is honest. Additional royalties will trickle in for the next year, then your game is done. Its life span has ended. If you want an income after that, you have to write another game.

SCO's claims regarding the Unix System V code from the mid 80's, a decade older than the most recent versions of DOS, ignore twenty years of Moore's Law depreciation. But the declining price each time the Unix technology changed hands (purchased by Novell, SCO/Tarantella, and SCO/Caldera respectively) matches Moore's Law depreciation fairly well.

SCO/Tarantella agreed to sell its Unix business to SCO/Caldera for a little over $53 million, although it eventually settled for around $36 million. [25]This is the combined purchase price of not just the historical AT&T codebase (including the versions up through System III, which they open-sourced), but also for the distribution/reseller channel, SCO's OpenServer/Xenix product line, and Novell's UnixWare.

An additional three years of Moore's Law depreciation would be two 18-month periods of 50% depreciation, yielding 25% of its original value. $53 million would become $13.25 million, and $36 million would become $9 million.

But only a small fraction of the purchase price was attributed by Caldera to the Unix technology. Their most recent annual report includes a breakdown of the Tarantella acquisition on page 46. They valued what they call "Existing technology (consisting primarily of UNIXWare and OpenServer)" at $5.8 million, and the "Distribution/reseller channel" at $26.7 million.[26] This means that they were willing to pay 4.6 times as much for the distribution channel as they were for the two Unix product lines the distribution channel was selling. (This is because at the time, they intended to repurpose the channel to sell Caldera OpenLinux.)

Of course what they were willing to pay for an asset, and how much they thought it was actually worth once they had it, are different numbers. On page 49 of that same annual report, SCO/Caldera notes that it wrote down half the value of its reseller channel, and two thirds of the value of the acquired Unix product lines. It explains the reason behind the writedown on the bottom of page 48:

Subsequent to the acquisition of certain assets and operations from Tarantella (see Note 3), the Company experienced significant unanticipated decreases in actual and forecasted revenue of the acquired operations, a significant decline in market valuations and general conditions, particularly in the information technology sector, a weakening of partner relationships, the loss of certain key executives, and other factors which indicated the recorded values of the long-lived assets were impaired.

This leaves the next table on page 49, which is still in use by SCO/Caldera's 2003 first quarter 10-Q filing, (on page 9, the table "Amortized intangible assets") [27] In this table, Caldera values the technology they acquired from Tarantella at a little under $1.7 million, with an estimated useful life of 5 years. (This is the period over which the asset will be fully depreciated for tax purposes.) The $1.7 million figure includes the value of the SCO OpenServer and SCO UnixWare product lines, as well as the historical AT&T codebase (most of which, specifically System III and earlier versions, they gave away for free by putting it up on their website for free download). Their own depreciation estimates have already brought that $1.7 million down to around 1.1 million since the purchase.

So either SCO/Caldera was lying to the Securities and Exchange Commission and the Internal Revenue Service, or at the time they filed the lawsuit they believed their interest in all the Unix technology they purchased from Tarantella to be worth just over $1 million.

This low valuation may in part be due to the fact they do not have free and clear ownership of this technology. On page 34 of their most recent annual report (filed at the start of 2003), they show a "Royalty payable to Novell, Inc." of just under $1.9 million in 2001, and another $1.4 million in 2002. More detail is available from the 10-k on page 42 (Restricted Cash and Royalty Payable to Novell, Inc.) and page 44 (Royalty Costs).

But more to the point, the main target of the acquisition was the distribution channel, not the Unix products (which were redundant, since Caldera already had a competing Linux product). In the same asset amortization table, SCO/Caldera also assigns a dollar value of $11.6 million to the distribution/reseller channel acquired from SCO/Tarantella; almost seven times as much as the acquired technology. Therefore, by Caldera's own SEC filings, their purchase of SCO's server division was mostly about acquiring a reseller network, and the Unix technology (including System V) was a side issue worth (at most) a little under 1/7th of the purchase price. (This is up from the 4.6 times as much they paid for the distribution network vs the acquired technology: as time went on, the reseller network became relatively more important to them.)

This matches numerous statements Caldera made at the time of the SCO acquisition, that the reason for the acquisition was to acquire a reseller channel through which to push their Linux products. We talked about this in the introduction to this rebuttal.

SCO/Caldera's lifespan estimate for the acquired distribution/reseller channel was 5 years. This may have been optimistic, since one of the more pronounced effects of their massive cost-cutting campaign, in addition to gutting their research and development effort, has been to dismantle the distribution/reseller channel they acquired. They've laid off personell, closed plants, and terminated business relationships. Page 11 of SCO/Caldera's first quarter 2003 SEC filing is devoted to their "restructuring plans", including "reduction in facilities", "severance costs of affected employees", and the liquidation of their United Kingdom subsidiary, "SCO Group, Ltd.".

The relevance of Moore's Law for this case is threefold. First, it took no special insight to foresee that Intel chips would become sufficiently powerful to support Unix. Secondly, that the reason Intel's x86 line came to dominate over competing processor manufacturers was due to IBM's use of x86 in the IBM PC, which became the widely cloned industry standard commodity hardware platform, and not simply due to Moore's Law as SCO states.

Third, the market value of the ancient Unix sources had depreciated to irrelevance at least a decade ago, which is why nobody outbid Caldera to purchase it, and damages significantly in excess of Caldera's purchase price for the technology (minus contined Moore's Law depreciation, 75% every three years) are completely unsupported. SCO/Caldera's own estimates from the start of this year, including OpenServer and UnixWare, place the technology's value at just over $1 million. Moore's law indicates that a computer technology purchased for $5.8 million should depreciate to a little under 1.5 million in three years due to the natural progression of the industry, without any help from IBM. This is roughly consistent with the observed behavior.

D. The Itanic Disaster

SCO purports to believe that the collapse of its business is due to malign meddling by IBM. Among other things, SCO blames IBM for the failure of the Monterey project, depriving SCO of a viable new proprietary Unix product to replace its aging OpenServer and UnixWare lines, perhaps even allowing SCO to attract the high-end enterprise customers its own efforts had failed to impress. But insofar as SCO is a victim of anything but its own incompetence, it is a casualty of one of the most expensive failures in the history of the semiconductor industry: the Itanium.

This appendix has three parts. The first explains what motivated Intel to produce the Itanium, what kind of industry transition Intel was trying to accomplish, and describes a similar historical transition Intel managed a decade earlier. The second and third parts explain why the Itanium project, and thus Monterey, failed so dismally.

Why Itanium?

Itanium was a joint project by Intel and Hewlett Packard (HP) to design a 64-bit processor to become the successor to the 32-bit x86 processors. It has failed spectacularly, garnering the nickname "Itanic". To understand what went wrong, a little history is required for context:

The original IBM PC, which was announced on August 12, 1981 used Intel's first 16-bit processor, the 8086, which could only access one megabyte of Random Access computer Memory ("RAM"). This is the processor Microsoft hired SCO to port Xenix to, and the main source of the "design challenges" they talked about in paragraph 36. (An early successor of the 8086, the 80286, could access 16 megabytes but only through the use of awkward and clumsy extensions which were of no use to most programs.)

The original mid-1981 IBM PC came installed with anywhere from 16k to 64k of RAM, depending on how much the purchaser wanted to pay. Moore's Law says that memory capacity at a given price point doubles every 18 months, so a simple mathematical formula can express approximately how much memory a high-end mainstream PC could be expected to have in any given year. [28] Every third year comes out evenly, so here is a table of answers for every third year, answers which match historical reality fairly well:

YearMemory Size
19808-32 kilobytes
198332-128 kilobytes
1986128-512 kilobytes
1989512-2048 kilobytes
19922-8 megabytes
19958-32 megabytes
199832-128 megabytes
2001128-512 megabytes
2004512-2048 megabytes
20072-8 gigabytes
20108-32 gigabytes

This is approximately the memory range on PC systems avilable for retail sale. After being sold the systems remain in use for a number of years, as part of the "installed base" for which new software could still be sold. Three years past the low-end is a good first estimate for the trailing edge of the installed base[29]. By using Moore's Law, Intel could easily and accurately predict that by the early 1990s the 1 megabyte of memory their 16-bit processors could provide to 16-bit DOS programs would fall off the bottom of the capability range that mainstream users were demanding in order to run then-modern software.

To prepare for this, Intel introduced its first 32-bit processor, the Intel 80386 (abbreviated "386"), on October 17, 1985. The 386 was plagued with initial technical difficulties and was not produced in large quantities the first year. 32-bit processors did not become the standard PC processor (with more than 50% of installed base using 32-bit processors instead of 16-bit) until the early 90's.

Intel knew that the switchover wouldn't be instantaneous, and prepared for a gradual transition by designing their new 32-bit processors with the ability to emulate the earlier 16-bit processors, in order to run the existing 16-bit software. In fact, the new 32-bit processors were faster at running the old 16-bit programs than the old 16-bit processors had been, so mainstream users had an incentive to buy a new 386-based system to run their existing 16-bit software when they upgraded their systems. Gradually, as the 386 and its successors became widely deployed, 32-bit applications (capable of using the extra memory but also requiring the newer processor type to run at all) trickled into the market.

The point of no return in the switch to 32-bit software was the release of Windows 95 on August 24, 1995. Windows 95 was the first operating system from Microsoft which required a 32-bit processor to run the OS itself. This came a full decade after the release of the 386.

The 386 was a good processor, which (with its successors) did eventually take over the PC market. The initial 386 could access 4 gigabytes of RAM. (Some of the 386's successors in the "Pentium" line can access up to 64 gigabytes (2^36 bytes) of memory through another set of awkward and clumsy extensions like those of the 286, but again this doesn't help most programs.) 4 gigabytes is still not quite a mainstream limiting factor today, but according to the table above it will become one for the high end of mainstream systems somewhere around 2005. Just as the move from 16 to 32-bit computing had been driven by the need to let programs access more than one megabyte of memory, the need for Intel's new "64" bit processor was rooted in the looming 4 gigabyte barrier.

To prepare for this, Intel teamed up with Hewlett Packard (HP) during the 1990s to design a new 64-bit proccessor. This new processor was originally code-named "Merced", and rechristened "Itanium" by Intel's marketing department around the time Intel started actually selling them. 64-bit processors like Itanium can access so much memory (264 is 18,446,744,073,709,551,616, or 18 exabytes, or 18 billion gigabytes) that Moore's Law may hit physical limits (such as the inability to make circuitry smaller than the atoms it's constructed from) before we have to switch again.

Why Itanium Failed

When Intel teamed up with other companies to create the 64-bit processor design which eventually became Itanium, they bit off more than they could chew. Instead of choosing to do a simple 64-bit extension of the x86 platform, they chose to do a whole new design based on the unproven "VLIW" (Very Long Instruction Word) technique pioneered by IBM in 1986. VLIW can be briefly described as a successor to RISC (Reduced Instruction Set Computing), which is a way of simplifying processor designs which IBM had previously invented. Intel's new project, originally code-named "Merced", was based on a very different design approach from traditional 16-bit and 32-bit x86 designs, and Intel and its partners had a lot of trouble making it work.

Even though Intel and HP announced their joint research and development project in June 1994, the release of Intel's new 64-bit processor didn't occur for more than five years after that. The release of new design was repeatedly delayed (see Problems delay Merced chip and Intel's Merced chip may slip further), to the point where partners HP and SGI revived their own RISC processor designs rather than wait for the joint effort with Intel to bear fruit.

On July 13, 1999, Intel announced that their 64-bit processor design (then still called Merced) had been completed and submitted to manufacturing, which normally means that actual chips would soon be produced and the processor could be expected to appear in finished computers about one year later. But more than a year later, the processor remained off the market.

On October 4, 1999, Intel's marketing department renamed Merced to "Itanium". More importantly, they announced a change of direction for the chip, abandoning the desktop in favor of the dot-com server market. This was immediately percieved as a horrendously stupid move (even before the dot-com crash of 2000). Without the economies of scale from tens of millions of desktop machines, the Itanium could never become as cheap as the existing 32-bit x86 processors, and would lose out on the all-important price to performance ratio that drives purchasing decisions in a vicious circle. Without millions of desktop users and the developers who service them, Itanium would never sprout the quantity and variety of supporting software the x86 market has. And beyond that, it would leave a market vacuum for somebody else to produce a 64-bit upgrade for the x86 for the desktop when the 4 gigabyte barrier struck home. That someone else would define the new processor standard, and reap all the advantages.

Without the desktop, Itanium could not be the next generation of x86. It became just one more bespoke 64-bit processor to run customized software behind locked doors in server rooms, competing in a very small market niche against established 64-bit versions of Alpha, PowerPC, UltraSparc, HP-PA...

Before the month was out, the nickname "Itanic" was coined. Its first public appearance was in an excellent analytical article about some of the poor technical choices inherent in the Itanium design[30]. The nickname quickly spread throughout the industry. A year after the Itanium was submitted to manufacturing, and yet was still not being manufactured, one journalist wrote, “Personally, I am tired of writing stories that the Merced Itanium is late every six months or so, only to be re-assured that it will arrive in the second half of next year or this year.”

The reason for the repeated delays, and Intel's retreat from the 64-bit desktop market, was that Itanium's performance was unacceptably slow. Due to its complexity, the part could not be manufactured at clock speeds greater than 800 mhz, at a time when 32-bit Athlon and Pentium processors were available at gigahertz (1000 mhz) or faster speeds. Worse, even at the same clock speed Itanium's performance was only around half that of a comparatively clocked 32-bit processor. The article Itanium 2 to Launch to Lukewarm Reception points out, "... enthusiasm over Itanium's once touted backward capability to run 32-bit Windows applications was largely doused last year after tests revealed that the chip could only process such programs about as fast as a 4-year old Pentium II chip."

The earlier seamless transition from 16-bit to 32-bit processors in the x86 line had occurred because the 386 ran 16-bit code faster than the original 16-bit processors could, and was fairly cheap (and became cheaper as compatible successors like the 486 and Pentium quickly followed it). In colloquial terms, upgrading to a 32-bit system was a no-brainer, and the bang for the buck was excellent. But although the Itanium could run 32-bit x86 software, it was both significantly slower and significantly more expensive (costing thousands of dollars for the processor alone, and over time that price went up while Moore's Law brought competing prices down). Intel was forced by competitor AMD (which made x86 compatible "Athlon" processors) to keep improving its 32-bit Pentium line, and the performance and cost effectiveness of the new Athlon and Pentium 4 processors left Itanium in the dust.

Intel struggled ahead, but by this point it was hard work merely to establish even a niche market for Itanium, and once vendors had real chips to test they quickly found numerous flaws. The article “Compaq: Itanium failed our server tests” notes that Itanium finally shipped commercially in May 2001, and quotes Dell senior vice president Joe Marengi as saying customer interest for Itanium servers was "effectively zero" by fall 2001. Only 500 Itanium servers had been sold by the end of 2001 (and another 2000 given away to seed the market). The average selling price for those 500 servers was $27,400 each.

Trying to stave off disaster, Intel promised for years that a second generation of the Itanium design (code-named McKinley) would raise the Itanic. This new version was launched as "Itanium 2" in mid 2002, but as that above article states:

While Intel has dominated the market for 32-bit processors, it failed to make much headway against 64-bit chips from Sun Microsystems and IBM with the first iteration of its Itanium line. Sun, IBM, and HP have produced 64-bit chips for years and have dominated the high-end server market with these products.

Despite the promises, Itanium 2 couldn't solve the fundamental design problems of Itanium. Only 3,500 Intel Itanium servers shipped in 2002. In August 2002, the research firm Garner estimated that Itanium sales would still be lagging behind other 64-bit chips (such as IBM's PowerPC, Sun's UltraSparc, and HP's Alpha and HP-PA lines) at least through 2007. As recently as December 2002, stories about Itanium 2 performance lagging behind Intel's own 32-bit Pentium 4 were still being written, but Itanium was by then such a disaster it was no longer particularly newsworthy. A recent technical editorial summarized the situation, "After some 10 years of development, Intel and HP have made a costly chip that seems suited to a few niches."

However, the need for more than 4 gigabytes of memory continued to press, leaving a market vacuum for a simple 64-bit extension to the 32-bit x86 design. The new processor should run existing 32-bit software quickly and cheaply, but allow new programs that make use of its extensions to access more memory. Since Intel refused to publicly admit defeat with Itanium, its main competitor AMD came out with its own 64-bit design called "Opteron" (which was code named "Hammer" during development the way Itanium had been code named Merced). Opteron processors started shipping April 22, 2003, and systems using them are already available from dozens of manufacturers. AMD is explicitly targeting both server and desktop markets, something a 64-bit upgrade of commodity x86 processors must be able to do to inherit the installed base of software. Unlike Intel, AMD is happy to compare Opteron's 32-bit performace against traditional 32-bit x86 processors. And while Itanium 2 systems cost around $10,000 per processor, the cheapest opteron processor as this is written costs $229, and AMD estimates a 4-processor Opteron server could be built for less than the price of a single Itanium. AMD has announced plans to discontinue production of its 32-bit Athlons and produce only 64-bit chips in the future, introducing even cheaper 64-bit x86 processors aimed at desktop and laptop systems.

From the start, AMD explicitly courted Linux developers to support its new 64-bit design, first releasing the technical specification for the new processor at LinuxWorld Expo in August 2000. Although Linux supports both processor types (along with over a dozen other architectures), as early as July 2002 Linus Torvalds was openly expressing his preference for AMD's 64 bit design over Intel's, and his hopes that Intel would follow AMD's lead. In Febuary 2003, Linus Torvalds strongly criticized Itanium, again expressing hopes that AMD's 64-bit extensions to x86 would become the industry standard. (All of this was before SCO surprised the industry by filing its lawsuit.)

IBM continues to hedge its bets. On the hardware side, it is supporting both Intel's Itanium 2 and AMD's Opteron, in addition to its own 64-bit Power PC processors. (This is similar to the way IBM simultaneously supported the Monterey partnership with SCO, the Trillian Linux on Itanium effort, and its own in-house AIX.) On the software side, IBM finally gave up on Itanium support back in February, before the lawsuit, saying, "Our view right now is that Itanium is like a science project. There's not a market for it."

Intel continues to plug away at Itanium, proposing to rip out the pathetic 32-bit x86 execution circuitry and fake it in software to bring the price down, and proposing several future generations of design revamps code-named Madison, Deerfield, and Montecito.

Between 1998 and 2003, the persistent failure of Itanium to gain any significant market traction slowly strangled operating system development projects targeted specifically for it, such as Monterey and Sun's Itanium port of Solaris.

For more reading on Itanium and the future of 64-bit x86, see IT Week's special coverage of the Itanium 2 launch, and eWeek's coverage of the Opteron launch. From a historical perspective, a good collection of the information available on Itanium at the end of 1998 is available at Merced Facts and Speculations.

The Monterey Fiasco

Note that the failure of Itanium was not the only reason for the failure of project Monterey. In the 1990's, the proprietary Unix market had once again fragmented itself into near-irrelevance with numerous incompatible versions, and seeking to salvage something from the mess by bringing together several vendors around a common standard was the traditional response. (See the response to paragraph 48 for some earlier examples of this traditional response, such as OSF/1). This unification is the reason IBM gave for launching Project Monterey as a multi-vendor effort including other vendors such as SCO. The article states:

The ultimate goal of the companies is to give software vendors a single platform to port to for 32- and 64-bit Unix systems on Intel and IBM Power microprocessors, they said.

Unfortunately, they wanted to create a single common Unix platform that did not include the dominant Unix vendors, Sun Microsystems and Hewlett Packard. Trying to create a cross-vendor standard Unix platform in 1998 without including Sun and HP was a lot like trying to create a multi-vendor standard Linux platform without including dominant players Red Hat and Debian, as SCO tried to do by joining UnitedLinux a few years later. Monterey predictably collapsed back into the next version of IBM's AIX, just as UnitedLinux is quickly becoming perceived as merely the next version of SuSE's Linux distribution. (Isn't it interesting the way SCO just keeps repeating the same mistakes?)

Predictably, Monterey was by no means the only "Unix on Itanium" effort. For a while Sun Microsystems planned to port its Solaris to Itanium, and lined up a number of large vendors behind it (including Fujitsu, Siemens, and NCR). IBM tried to compensate by recruiting other big companies like Samsung and Computer Associates to support Monterey. For a while, other companies (including SGI and Compaq) toyed with the idea of porting their own in-house Unix variants to Itanium, before deciding there was insufficient market justification. By the time Monterey ground to a halt, Sun's Solaris on Intel project project succumbed to inter-company rivalry between Intel and Sun's own "Sparc" processor division, and the general realization that Itanium wasn't ready for prime time and they were just throwing good money after bad.

Despite its spectacular failure to master 64-bit computing, Intel remained the industry's leading 32-bit processor vendor, and ignoring its efforts was uncomfortable for many corporate decision makers. Many of these companies decided the cheapest way to hedge their bets against a sudden turnaround in Intel's 64-bit fortunes was to support Linux on Itanium. Hobbyists maintain Linux support for a number of unusual types of hardware, including obsolete Vax minicomputers, or a number of long-discontinued minicomputers like Amiga, Atari, and early Macintosh machines. No matter how poorly Itanium did in the market, to the Linux crowd it was just one more oddball niche platform to support, no big deal. More than that, the Linux port to Itanium was ready by the time the prototype Itanium hardware was available, but the proprietary operating systems were not, and were publicly happy that Intel's repeated hardware delays gave them more time to work.

As the big players squabbled, Linux naturally and steadily gained ground in the hearts and minds of corporate users looking for a working 64-bit platform. If Intel's billions churned out a third or fourth revision of Itanium that miraculously became relevant, Linux would support that hardware. Meanwhile, Linux versions of 64-bit software could be run on any of the other 64-bit processors, to actually be deployed and in production while the Itanium drama played itself out.

A working version of Linux for the Itanium was already available before IBM cancelled project Monterey, and (due to repeated delays) before Intel even commercially shipped the first Itanium systems. IBM and SCO were quite aware of this project at the time. SCO/Tarantella's CEO even gave an amusing interview dismissing the Linux porting project, codenamed "Trillian", as irrelevant. They were more worried about the proprietary Unix market leaders, Sun and HP.

[1] Unix source code has been regularly used as a teaching tool in college operating systems courses ever since the inventor of Unix, Ken Thompson, took a year sabbatical from Bell Labs in 1975 to teach at his Alma Mater, the University of California at Berkeley. The students he left Unix code with went on to start the Berkeley Software Distribution, BSD. Andrew Tanenbaum created Minix to replace Unix source code in operating systems courses only when the AT&T divestiture made Unix source licenses prohibitively expensive for universities to afford. It's difficult to argue that trade secret status can apply to source code that has been licensed for use as training materials in accredited University courses.

[2] (The edgar archives are badly organized that far back, but you can go here and search for "EX-10.14" to find a complete copy of the GPL text. A helpful list of frequently asked questions and answers about the GPL is also available online from the GPL's authors, the Free Software Foundation.

[3] United States v. Western Electric Co., 1956 Trade Cas. (CCH) § 68,246 (D.N.J. 1956).

[4] In recent interviews with Byte magazine and Linux Journal, SCO's CEO Chris Sontag indicates that SCO is willing to go after BSD as well if they can get away with it. It's interesting to note the drift in SCO's lawsuit objectives: originally a contract dispute with IBM over code licensed during Project Monterey, then a complaint that Linux contains System V code from the 1980s, and now claims that BSD may be a target in future if they can work around the earlier settlement.

[5] The fact that AT&T had even tried to compete against its own Unix licensees left the same kind of bad taste in the mouths of Unix gurus that modern computer programmers feel today when Microsoft tries to leverage its OS monopoly against Windows application vendors like Stacker, Netscape, or Oracle. This was an important element of the "Unix Wars" between System V and BSD. One reason the Unix crowd is now leading the charge against Microsoft's OS monopoly is they had practice against an earlier monopolist — the antitrust trial calling for Microsoft's breakup a few years ago brought back a lot of memories for old-time Unix gurus.

[6] Of course Microsoft wasn't particularly visionary here, either. Paul Allen, the Microsoft co-founder who was technical enough to find Unix interesting, commissioned ports to both the Motorola 68000 and the Intel 8086. After Microsoft found out which of the processor lines IBM had chosen, the other version was discontinued. Xenix itself was deprecated at Microsoft after Paul Allen retired in 1983 for health reasons. This is how SCO wound up owning it.

[7] Yes, we know about the Compaq Deskpro 386, but it's not relevant to this discussion.

[8] Mark Taylor of IBM's WebSphere MQ Technical Strategy organization at the IBM Hursley Laboratory points out that individuals and smaller departments within IBM did take notice of Linux years before IBM's corporate executives became aware of it. Mark writes:

In fact, IBM released something much earlier, in February 1996. It was a client for the MQSeries product, running on a Linux 1.0 kernel. (I used an early Slackware distribution.) This was freely downloadable by the general public. One of the selling points of MQSeries has been its platform coverage, and Linux was one of the systems we could see becoming important even then, at least as a "client" operating system.

I remember visiting the DB2 development laboratory later that year, and being asked by their team whether it would be a good idea for them to also port DB2 to Linux.

Even before that, someone at IBM donated a Power PC system to Linus in 1994 so he could port Linux to that hardware platform, but Linus had already received a system based on the 64-bit Alpha processor from Jon "Maddog" Hall of Digital Equipment Corporation, and completely ignored the donated Power PC hardware.

All three of these events predated IBM's Linux kernel development efforts by a number of years. The Power PC was donated hardware (which was ignored). The MQSeries client was not a product, but a client for a product. (It was basically a building block that could be used to develop an application; a piece of middleware infrastructure to allow Linux clients to talk to other (non-Linux) systems through the network.) The DB2 port is important because it was the first complete runnable Linux application offered by IBM, actively publicized as such, and presumably produced with the blessings of upper management.

[9] Technically, although the Alpha port was the first one to get integrated into Linus's tree, and the code reorganization to accomodate the Alpha platform set the model for future additions of support for new platforms into Linux, it wasn't actually the first port. A Linux version for the Amiga, running the Motorola 68k processor, came first, but was never integrated into Linus's tree for reasons Linus describes in his chapter of the book Open Sources.

[10] A good resource for measuring linux-kernel development leading up to 2.4 is the Linux Kernel Version History table. To browse more recent versions, see Linux Headquarters' Kernel page.

[11] Modeling is where a digital artist uses a graphics workstation to create a 3D model, a bit like creating digital puppets, painting them, lighting them, describing how they move... Rendering is the process of running the digital puppets through their paces to crank out finished graphics images, producing a video one frame at a time. Video capture is the digital recording of a live video feed (this is what a Tivo does), and playback is the production of a live video signal from a computer file. Editing is splicing together snippets of video, cutting from shot to shot or scene to scene. Compositing is where two overlapping video streams are merged into one, including things like "blue screen" work, inserting computer generated images into live video footage, or more subtle video effects like making lightsabers glow or rayguns fire.

[12] See page 3 of the MPS 1.4 specification. There was a prototype document released in October 1993 under a different name, but MPS 1.1 was the first widely implemented SMP standard from Intel.

[13] IBM couldn't sell any NEW machines during this time, but rather than lay off workers it kept making them and filled up warehouses with unsold tabulators, which turned out to be a great asset when F.D.R. started the Social Security administration and the federal government bought IBM's entire backlog of tabulators and ordered a bunch more on top of that. This has nothing to do with SCO, but this is the kind of footnote you get in documents written by a historian. :)

[14] The lead author of this document, Rob Landley, used to hang out in his office and listen to him bitch about Zip disks and how SCSI controllers lie about drive geometry and were never intended for use with removable media. Fun guy, I learned a lot. I met Shiela Harnett as well, she was also a senior OS/2 developer. I haven't seen her since she went on to become technical lead for the IBM Linux Technology Center, sometime after leaving OS/2 to work on JavaOS.

[15] SCO/Caldera's most recent annual report, page 10, employees: "As of October 31, 2002, we had a total of 340 employees. Of the total employees, 75 were in software engineering, 113 in sales, 20 in marketing, 51 in customer service and technical support, 20 in customer delivery, and 61 in administration."

[16] "Virgin programmers" is an industry term of art meaning programmers who have not seen proprietary code, and are thus capable of reinventing the wheel without being accused of plagiarism. SCO isn't the first company to sue people for reusing the obvious, and the easiest way to prove how obvious something was is to do a clean room clone where the developers have nothing to go on but a description of the desired result, a descripion which a lawyer has confirmed is not anybody's proprietary property. The term was probably invented by Compaq, which cloned IBM's BIOS in the early 1980's to produce the first 100% compatable IBM PC clone. A clean room clone using virgin programmers (often hired straight out of college or even high school to make sure they've never seen anything proprietary) is something you do if you expect to get sued anyway, and want a paper trail to prove your case in court. It's faster and cheaper to just use experienced programmers.

[17] The SCO purchase from Novell was mostly a stock swap and profit sharing plan, so it's difficult to pin a precise dollar value on it. Novell's 1996 annual report sums up the transaction as follows:

On December 6, 1995, Novell completed the sale of its UnixWare product line to the Santa Cruz Operation, Inc. (SCO). The Company expects to report a gain on this transaction in the first quarter of fiscal 1996. Under the agreement, Novell received approximately 6.1 million shares of SCO common stock, resulting in an ownership position of approximately 17% of the outstanding SCO common stock. The agreement also calls for Novell to receive a revenue stream from SCO based on revenue performance of the purchased UnixWare product line. This revenue stream is not to exceed $84 million net present value, and will end by the year 2002. In addition, Novell will continue to receive revenue from existing licenses for older versions of UNIX System source code.

Novell later put the cash value of the sale of the Unix technology at $19 million, plus ongoing revenue from Unix royalties (royalties which SCO/Caldera continues to pay to Novell). These royalties were lumped together with "Service, training, consulting, and other". The relevant description from page 11 of Novell's 1999 annual report is:

Service, training, consulting, and other includes revenue from customer service, training products and courses, consulting for network solutions, UNIX royalties, and other. These revenues were $166 million in fiscal 1998 compared to $161 million in fiscal 1997, and $230 million in fiscal 1996. The increase from fiscal 1997 to 1998 was due to the growth of service revenue more than offsetting declines in UNIX royalties and other. The decline from fiscal 1996 to 1997 is mainly the result of declining sales of UNIX licenses as well as a one-time $19 million paid-up royalty recognized on the sale of UNIX technology to SCO in fiscal 1996. Service, training, consulting, and other revenues were 15% of revenue in fiscal 1998 compared to 16% of revenue in fiscal 1997, and 16% in fiscal 1996.

The 72.4 million estimate quoted earlier is as good an estimate as any we've found. Novell wanted to wash its hands of Unix, and wasn't particularly picky about how it did it. The true figure is almost certainly less than $100 million.

[18] Pull out your calculator: if you start with a billion dollars and cut it in half every 1.5 years, and do that for twenty years, you get 1,000,000,000/(2^(20/1.5)), which is 96,887. Multi-million dollar mainframes often get thrown out after ten or fifteen years if they aren't constantly upgraded, now you know why.

[19] It does repeatedly express concerns that Sun Microsystems had announced plans to open-source Solaris. Bottom of page 14, for example. Sun appears not to have followed through, possibly due to poor community acceptance of the Sun Community Source License.

[20] Here's what Caldera's own SEC filing of a question and answer session about the SCO purchase had to say on the matter:

12. How does this help Caldera, Inc. competitively?

Caldera, Inc. will be able to make the most of the synergies between the two existing companies, resulting in:

  • World's largest channel,

  • Increased breadth of product options and accelerate Linux technology plans,

  • OEM's can now get scalability of product, professional consulting and engineering services and instant global marketplace,

  • World-class global services and support infrastructure

You could also read about it (with quotes from Caldera's CEO) in an article from Computer Weekly, read a good analysis from 2000 about Caldera's acquisition of SCO from the register, or a sceptical take from computerworld.

[21] For more on this, see SCO/Caldera's most recent annual report. UnitedLinux shows up in many places (search for UnitedLinux), but the bit at the bottom of page 63 is where the ownership stake and supporting payments are detailed.

[22] According to Microsoft's quarterly report for the relevant period, there were 5,332,337,924 shares of MSFT outstanding, against which they took a 3 cent per share charge to pay for the Caldera settlement, which would be $159,970,137.72, or just under $160 million.

[23] According to the article, the shell game apparently went something like this: Caldera was split into three companies: Caldera Systems (a Linux business run by Ransom Love), and an embedded systems company quickly renamed Lineo, were both spun off from the original "Caldera Inc.", leaving it an empty shell with no employees, run by Bryan Sparks (also CEO of Lineo). Caldera Inc. received the Microsoft suit settlement money on behalf of Canopy, then faded away. Caldera Systems went on to have an IPO in order to fund its operations (presumably deluding some investors into believing it had received the Microsoft money, and thus was in much better financial shape than it really was.) Shortly thereafter it switched corporate identity again to "Caldera International". (And you wonder why it forgot it had been a Linux company...)

Canopy sold Lineo to Motorola in 2002. Bryan Sparks went on to run Device Logics, another Canopy company based on DR-DOS, a technology once owned by Caldera, then Lineo, but never relinquished by Canopy. Another Caldera technology taken away from Caldera and used to start another Canopy company was Caldera Volution manager, used to form Canopy's holding Volution. (If SCO does come up with any major cash windfall, the authors of this commentary would not be at all surprised to see Canopy find some way to absorb it again.)

At the time of the lawsuit's filing, Ray Noorda owned 47.7 percent of SCO/Caldera's's stock through the Canopy Group, a division of the Noorda Family Trust which Ray Noorda still personally oversaw as recently as 2001. Although Canopy has reduced its ownership position slightly since the lawsuit (taking advantage of the increased stock price to cash out), it still owns more than 40% of Caldera's shares. According to page 66 of SCO/Caldera's most recent annual report, SCO's chairman of the board is the president and CEO of Canopy (meaning he works for Ray Noorda the way Darl McBride works for him). SCO/Caldera leases its office space from Canopy. On page 67 we learn that SCO's largest creditor is the Canopy Group.

Presumably, Canopy could stop the lawsuit with a phone call, but hopes to profit from it instead.

[24] A number of interviews with Gordon Moore are available online, including a 1995 interview with Stanford University and a 1997 interview with Scientific American.

[25] According to this 10Q from Tarantella (Note 7 beginning page 14), Caldera paid SCO a total of $23 million in cash (the oft-quoted figure of $7 million in Q1 2001 was just the first installment), plus stock totalling 28% of the ownership of Caldera, and an $8 million prommisory note to be paid in installments. Tarantella recorded $53,267,000 as the sale price of server and professional services groups, but in reality they didn't get that much. Caldera's financial position rapidly deteriorated, their stock price collapsed, and Caldera announced a 1 for 4 reverse stock split (a desperation move to avoid being de-listed by their stock exchange). This convinced Tarantella it would be a good idea to close out their debt and equity positions with Caldera as quickly as possible, and get what they could out of it. So in the second quarter of 2001, Tarantella allowed Caldera to cash in the $8 million promissory note for a single up-front cash payment of $5 million, and convinced Caldera to buy back $555 thousand worth of their stock. In the third quarter, Tarantella sold $4,360,938 of Caldera stock on the open market, and Caldera bought back the remaining $3,059,250 worth. By September 30, 2002, Tarantella had unloaded all its Caldera stock. Add in another $445k of miscelaneous payments from Caldera to Tarantella, and the total payment Tarantella received for its Unix division was (445+555+5000+4361+3059=13,420 thousand, plus the original $23 million cash, equals) just under $36.5 million. This matches Forbes' estimate of around $36 million.

[26] The $93.8 million figure they use here as the price for the acquisition was based on an extremely optimistic valuation of the stock they paid with, and is basically cancelled out by the $66 million of "Goodwill", which is an accountant's way of saying "we paid more than we can justify on paper, and get to depreciate this imaginary substance for tax purposes". It's all accounting games, just ignore it. The ratio of the figures to each other is what's important, and can be normalized against the real numbers later.)

[27] Note that these are their valuations and lifespans for actual assets, not accounting tricks like goodwill. The paragraphs above the table under (4) INTANGIBLE ASSETS state "Under SFAS No. 142, goodwill and other intangible assets with infinite lives are no longer amortized, but rather are assessed annually for impairment... Upon adoption of SFAS No. 142, the Company reassessed the useful lives of its intangible assets and related fair values and reduced the carrying value of its goodwill to $0."

[28] Assuming memory capacity doubles every 1.5 years and starts with a range of 16-64k in mid-1981, the formula to find the high end of the range for a given year would be "2(5+((year-1980)/1.5))". Plugging in 1981 and a half for the year yields 2 to the power of 5+(1.5/1.5), or 2^6, which is 64. This answer is the kilobytes of memory a high-end system could be expected to have in that year, with the low end equipped with 1/4 as much memory.

[29] The reality is somewhat more complex, including upgrades of existing systems and corporate decisions to leave alone anything that works, no matter how obsolete. This level of detail is beyond the scope of this discussion.

[30] Dr. Robert J. Harley points out that the true origin of the name Itanic seems to have been in the user comments appended to an October 4, 1999 Slashdot article about the name change from Merced to Itanium, where the name "Itanic" spontaneously arose several times. It was passed along on Usenet the next day, and found its way into the Low End Mac article on the 20th, where it entered the mainstream.