OK, I found draft-levinson-cid-01.txt document. It looks pretty good.
Comments:
I think you need to be more careful about your description of the
handling of %xx hex-encoding of characters in content-ID tags that are
not otherwise representable inside URLs.
To transform a cidurl or
midurl into a valid content-id or message-id, surround the
id-spec part with the enclosing brackets, i.e.,
but in fact you also have to decode any URL-encoded sequences %nn. It
seems that these URL schemes aren't really *subsets* but *encodings*
of content- and message- IDs.
Why, in fact, are two schemes needed? Can the same id-spec appear as
both a cid and a mid?
Is the type of the object referred to by a mid always implicitly
message/rfc822?