I'm very sympathetic to the substance of Harald's about being a little
more specific about gateways removing the Content-MD5 field. I think his
reasoning should be carried a half-step further.
We've been trying to establish the principle, e.g., with the SMTP 8bit
extension doc, that gateways should document, in Received fields (comments
if necessary) any transformations they apply to messages other than the "just
receive and hand off" behavior of a relay. Consistent with that, it would
be better to specify that, if the Content-MD5 field was removed, a
Received header (or equivalent in the target mail environment) be inserted
that shows not only the receipt of the message, but that this field was
removed and optionally why.
Question for Marshall: While my personal preference is to get this type of
guidance/specification out contemporaneously with the specification document,
I can see a strong case for an (unfortunately not very modular) standards-
trace "MIME and gateways" doc that, among other things, works its way
through a number of these "what if that doesn't mean anything on the other
side" issues. Thoughts?