BTW, I just noticed that my last message was sent to ietf-smtp, not
ietf-822. Sorry about that, but I was just replying. If you want that
message (it's on the "what's the part in multipart" issue with message
id <9104272118(_dot_)AA10520(_at_)polya(_dot_)Eng(_dot_)Sun(_dot_)COM>), please
email me and I'll
forward you a copy.
On to business:
It occured to me as I was making lunch earlier, that some people might
consider digests and multipart messages to be the same thing and, so,
there should be one mechanism for both.
The problem is that digests are collections of messages while a
multipart message body is a collection arbitrary kinds of things.
Digests might have text at the beginning and end, while multipart
messages don't (although I could see some debate about this). At best,
a digest is a special kind of multipart -- where all the parts are
messages.
The difference between the RFC-XXX approach and the approach that I like
is how one tells the difference between data and message (that contains
data). In both schemes, a g3fax part looks like
--- zxcvb
Content-Type: g3fax
... fax data ...
--- zxcvb
(there would probably be a Content-Encoding also, but let's keep it
simple). Now consider a part that is a message containing a g3fax. In
RFC-XXX it looks like
--- zxcvb
From: ...
To: ...
Subject: ...
Content-Type: g3fax
... fax data ...
--- zxcvb
Although the Content-Type is "g3fax", this is really a message
containing fax -- the way to tell is to notice the presence of the 822
headers ("From:", "To:" and "Subject:").
In the scheme that I like, the part looks like
--- zxcvb
Content-Type: Message
From: ...
To: ...
Subject: ...
Content-Type: g3fax
... fax data ...
--- zxcvb
The Content-Type part header indicates that what follows is a message.
The inner Content-Type indicates what the body of the message is
(g3fax).
Clearly either scheme works. But, in my humble opinion, the second is
cleaner and a whole lot easier to implement.
I'd support the introduction of a "digest" Content-Type that uses the
same separation syntax as "multipart". With "digest", it is known that
each part is a message, so "Content-Type: message\n\n" would not appear
at the beginning of each part. I'm against allowing in a "multipart"
body, text before the first part and after the last. However, the
"digest" type could allow this if people think it's necessary.
Pete