On Thu, 4 Mar 1993 11:44:02 PST, Bill Janssen
<janssen(_at_)parc(_dot_)xerox(_dot_)com> said:
Bill> Perhaps we should just write another RFC specifying a
Bill> number of different variants of multipart (including the
Bill> two attachment models), and register them and start
Bill> using them. I have really become convinced that this is
Bill> tantamount to designing a compound document format,
Bill> though. Big job.
Big job indeed. I really think this is the wrong approach; compound
documents should be application subtypes, using the Content-ID of
other bodyparts to reference them. Trying to make MIME provide this
functionality by playing with a nightmarish array of multipart
subclasses seems to me a good way to confuse implementors, introduce
unneeded complexity, and insure early obsolescence!
If we instead adopt the approach with the Content-Disposition header,
we get seamless backward compatibility, a clean separation of typing
and disposition info (I admit that multipart/parallel is an anomaly,
but it would seem we are stuck with it, and it's semantics are pretty
well understood) and a documented, extensible mechanism for
attachments.
I already have most of a draft of this done, and I'll mail it to the
list tomorrow, most likely.
-Rens