Ned Freed writes:
So, what header-set helps the mailer with is to make some more
intelligent message to the user than multipart/whatever which
is treated as multipart/mixed.
I'm afraid I don't see the point. You have now argued, quite successfully as
far as I can see, that it is dangerous for a reader to treat
multipart/header-set with an unrecognized initial header part in any special
way. In other words, it should be treated as multipart/mixed if the header set
isn't known. But how does this differ from handling of multipart/whatever
I might have to rewrite the sentence once more. As you know english is
not my first language (it's swedish 8^) so:
That is exactly my point. I can (still) not see what multipart/header-set is
"better" than multipart/foo. The only difference is that IF you use
header-set, you can hint the user about the fact that he really need
to have knowledge about the first part, otherwise it's just a header
which is followed by the actual data. The header can NOT be skipped.
It must be treated as multipart/mixed.
So, then you have a known multipart type, header-set, which has
a sub-part which is unknown, then the multipart message should be
treated as a multipart/mixed message. This is a very strange
way of doing things I think.
I thought I should be convinced to use header-set, and like it,
see the benefits, but when reading all messages on this list,
I now more and more like multipart/foo instead.
Until and unless I see some tangible reason why the semantics of a header-set
are useful to someone, somewhere, I will continue to see no point in this
proposal. I can live with it with the added parameter, but I still don't think
it is useful.