Dave, gang:
Dave is pointing at something that worried me also about Multipart/PEM:
That we have pieces of "comprehensible" stuff together with something
application-specific, inside an application-specific header.
Why not define an extendible mechanism for Multipart:
content-type=multipart/structured; control=first [or last]
where we then have the semantics that:
- A body part apart from the control has its normal meaning
- The control body part has content that is about other parts.
So, if you cannot parse the control, you can look at the other parts, but
if you can parse the control, but not the other parts, you might as
well junk the whole thing.
So, I can read signed PEM messages (or PCR, or whatever) without having
to implement PEM or PCR; I lose the check, but not the messages. In the
same way, I get to read MAC GIF pictures without knowing what a MAC is.
(No, I did not suggest it for structuring Attachments. Others should
get their fingers burned.....)
Thoughts?
Harald Tveit Alvestrand