I suggest a comma-separated list of subtypes if there is an alternative
in the header-set's first body part.
No objection from this quarter.
As long as we're going to commit to having the subtypes of the header-set match
what's in the multipart parameter, I have no problem with this either. I still
question whether or not requiring the parameter to be the same as the header
set type isn't overly restrictive in the long term. But then again, I would
much rather flush all these proposals completely and go back to multipart/foo,
so who am I to say what relation should exist?
Ned