Keith,
Tree-pruning seems like a good idea (especially in a paperless
society?) but please note that Sender-MHS and Receiver-MHS can't
be stuffed down into SMTP, since there may be any number of
(email) hops between the participants, possibly none with SMTP.
Further, one could easily model end-to-end reliability mechanisms in
much the same way as TCP provides. This is below the application
level, but is not "in the network". That is, the two end systems
participate, but as part of a coherent infrastructure for applications,
rather than as part of the fully-integrated, underlying, network
service.
So, Sender-Receiver focus strikes me as an excellent, initial focus,
but should this be to the exclusion of having any integrated framework
for encoding the end-to-end protocol, other than the raw primitives
of MIME?
(Note to other list members: Keith and I chatted about some of these
issues, last week, privately, so I would VERY much like to hear some
opinions from other group members. It is, by now, painfully obvious that
I think the issue reasonably important, but if there is no real interest
by the group in pursuing it, then I'm disinclined to push the
point single-handed.)
Dave