As I understand it the BNF, there can be multiple resource-refs separated by
commas. So I would claim that every bit of information that is present in
the RFC-XXXX syntax would be representable in the simplified syntax.
This leads me to wonder. If we really *must* have versions, why *must* we
have the version for every content type that has a resource ref? This is a
flaw in the syntax that becomes glaring as soon as we fold all the textual
types together. Why can't the versions become a resource ref? For example:
Content-Type: BINARY;U-LAW,VERSION=1.0.1, ...
Again, I think it's important that the top-level content type be severely
restricted (and preferably not open-ended at all) and the open-ended stuff
moved to the resource refs.
At the top-level, there are three major paths:
. stuff that is reasonable to present to a human
. stuff that is reasonable to present to a program
. stuff that encapsulates other stuff
What you're telling me is that the default is the second, but nobody is going
to go out and say that, and sooner or later some clown is going to put
something there that I don't recognize that I make the wrong choice with.
I am ameniable to picking a name other than "BINARY", given that I consider
things such as PostScript to be "BINARY".
Please also reconsider mandatory versioning.