Dan Kohn has written an excellent justification for this effort.
We seem to be approaching agreement on a specification for a subset of HTML that can used as an RFC distribution format and be rendered into flat-ASCII. I have developed a draft specification for this subset, RFC-HTML, and written a Perl program which compiles the specification file into flex code for a validator.
I have written another flex program which demonstrates that it possible for the RFC editor to automatically massage "naked" RFC references in any submission format into URLs.
You can browse the draft RFC-HTML specification or download an archive containing the specification and the programs. Comments, criticisms, and improvement are solicited. I also keep a change log so you can see how the pieces are evolving.
The next step, once we've settled on a suitable subset of HTML, is to describe a way to embed any necessary meta-information in HTML so that the traditional flat-text representation can be generated from the HTML. And to write the simple Perl program (using lynx -dump) to do the generation.
(Another possibility -- possibly a better one -- would be to adopt a system like SGML-tools for RFC markup.)
At that point, RFC 1543 could be amended to permit (or even require) HTML or SGML in the designated subset as the master form for RFC submissions.
This effort is not an official IETF project, but the discussion group (which spun off the DRUMS mailing list) includes some IETF heavyweights. It may result in a proposal to the RFC editor.
There is a mailing list for this group, rfcs-in-html@cs.utk.edu. You can subscribe in the usual way by sending mail to rfcs-in-html-REQUEST@cs.utk.edu. There are web-accessible mailing list archives.
You can read the mailing list's issues page describing the state of debate and what we think we've figured out.
Eric S. Raymond <esr@snark.thyrsus.com>