Actually the only way you can assign locally-unique IDs to a forwarded
message without scanning the forwarded message is to use globally
unique IDs.
If we want locally-unique IDs, we have to define what "local" means.
Current message? Current level of current message? Current message, but
not going into message/rfc822 body parts? What about message/external-body
where the external-body is of type multipart/mixed? Can we refer to
body parts inside that part of this message? Can we do it if some
mechanism "pulls in" the message?
Actually, I think Content-ID: should be dropped from the spec.
So far, no application is defined that uses it, and placeholders
seem to be not so neat an idea.
Opinions?
(Marshall, why do you cache body parts? Why not use message-id + number?
Content-id was never mandatory; my PP gateway will not make them; what
will you do then?)
Harald A