I'm afraid I was derelict in my duties and hadn't read the draft until
your recent mail prodded me to do so. Basically, I think it's fine. I
agree with the suggestion of moving towards more "standard" 822-style
BNF notation for the grammar, but that's a minor point.
I can understand why Keith is uncomfortable with the "difference"
parameter for multipart/alternative, but I think it is actually a very
good model. We might want to add the restriction that new values for
"difference" should always be the names of Content-* headers (the
current document just says they should be the names of headers). That's
because there's already the general rule that, inside a multipart, the
Content-* headers are the ones that have semantics that gateways should
try to preserve if possible.
Keith's suggestion for a general Content-attribute header is intriguing,
but sound potentially complex. Keith, do you care to elaborate the
details of how such a thing would work? In the absence of a very clear
alternative, I'm pretty happy with the "difference" extension to
multipart/alternative. (Although it does provide yet another bit of
evidence that my implementation strategy of building
multipart/alternative INTO the metamail program, as opposed to using an
external viewing program was totally wrongheaded and braindamaged. But
I already knew that, sigh...) -- Nathaniel