We could either remove the
specific grammatical entities for text-type, etc. -- probably my
preference -- or we could add a more complete grammar, along the lines
I'd go for the former, too.
I've not looked too closely at the grammar before, but unless I'm
being stupid, there are *two* formal syntaxes for the field-body of
`Content-Type': one deriving from the `content' token, the other from
`text-type' and friends. The definitions overlap, but are inconsistent.
I assume the rationale for `text-type' et al is to forbid `text/jpeg'
and other such nonsenses. I think it's a mistake to try to use EBNF
for this: it's just not powerful enough to do it sensibly (anyone fancy
defining a vWG for 822/MIME? :-)
Even if you could do it, I can't imagine anyone implementing it that
way. It just makes it harder to use the formal grammar at all (as the
starting point for a yacc-based parser, say).