This has been a most enlightening discussion. I will attempt to
summarize my perceptions of it:
1. Not all that many people found "<nl>" unacceptably ugly, I was
probably unduly influenced by the ones who had said they did because
nobody else was speaking up. Moroeover, if my generator was smarter
(using <paragraph> liberally) it wouldn't even be as ugly as it is.
2. Solution #2, the "newlines -1" solution, is preferred by many
people, and might have been a better idea, but opens up new cans of
worms vis-a-vis SGML and separation of richtext to a new document.
Given these facts, I would like to withdraw my suggestion. Personally,
I believe that any minor improvement that solution #2 might represent is
far less important than the major disimprovement that would result from
moving richtext to a separate rfc. (Readers of ietf-822 might not be
aware of this, but so far it is richtext in particular that has
generated the most enthusiasm among early users of the metamail
software, largely because it is so easy to generate even from character
terminals.)
I have always maintained that the best goal for richtext was to make the
email community realize that enriched text in the mail was so valuable
that an even "richer" text was required. In this regard, I've long
argued that richtext doesn't have to be perfect, just good enough to
make that point, because if it makes that point it will inevitably be
superseded by a richertext some day. This discussion has simply
convinced me that we've crossed the line and that any perceived
improvements now belong to richertext rather than richtext, i.e. that
the definition of richtext should be set in stone as-is.