I have no real objection to making richmail syntactically compatible
with SGML, though it would have to be an extremely small subset. (I'm
also not qualified to judge its compatibility, I think.) But I don't
see any real advantage to doing so, either: the act of defining the
subset will be no less cumbersome than defining a new syntax, and I
can't think of any circumstances in which the compatibility would really
be useful, given the semantics-free nature of any SGML parser. I'm not
opposed to the idea, merely skeptical that it makes a difference.