I mostly agree with your comments.
But i disagree with:
Section 2.1 refers to the MHS as monolithic; it is not --
it is composed of autonomous agents (MTAs, MSAs, MDAs,
gateways, list expanders) and a user may receive a delivery
notification of some sort from any of those agents.
For users the (image of the) MHS is monolithic.
In This section I would keep it that way.
Your respons. made me also think about the MTA - MDA distinction.
IN RFC2821 there is no MDA.
I think the MDA is a new entity first mentioned in this draft.
The closest RFC reference I could find are the
Local delivery agent and Local gateway agent in RFC 2033 - Local Mail Transfer
Protocol.
I like the concept of MDA and want to distinguish it more clearly from the MTA.
I would suggest the MTA only works with domain part of the message, the MDA
with domain parts and local parts.
So the MTA does the transfering, the MDA does all the rest.
If an operation needs the localpart of the address it is a MDA function.
(auto responders, mailinglist actions, the SMTP VRFY, EXPN, ETRN commands ect.)
Architectonal ideas
Eexcept with LMDP, an MDA is always hiding behind an MTA.
LMDP uses SMTP-Direct (sending to a fixed ip address, not to a public MX record)
There is no direct contact between MTA outside the administrative domain and
the MDA in that administrative domain.
There should be a chapter about the RFC2821 MTA
how in this draft it has architectionaly been split up in
MSA for submission
MTA for transfer
MDA for message "delivery"
and the
MS for storing
Another point the administrative domain.
Administrative domains can be nested / overlapping ed.
Within one administrative domain globaly valid MX records are NOT used.
MX records are used for transfer between seperated administrative domains.
within an Administrative domain SMPT-direct. (sending to only one MTA) is used.
What is your standpoint about Users in general?