Nathaniel, I agree completely. In fact, this has been a problem when
trying to be compatible between the Andrew mime mailer (where all
body parts are conceptually "concatenated", and Sun's attachment
model (where body parts are conceptually distinct, separate
documents).
I believe that both are valid models; MIME shouldn't take a position
that discourages either one. But we absolutely need a way to differentiate
which model was intended.
I tried to make this point in private correspondence a couple of months
back, but I never communicated clearly enough to make my point.
Neil
From: Nathaniel Borenstein <nsb(_at_)thumper(_dot_)bellcore(_dot_)com>
To: info-metamail(_at_)thumper(_dot_)bellcore(_dot_)com
Subject: Re: file attachments in MIME
I've also been treating application/octet-stream as a generic
attachment. What some of us now think might be needed is a way of
tagging almost any content-type as being semantically intended as an
attachment rather than an integrated piece of the message. I'm not yet
sure what the right answer is... -- Nathaniel