Bart Schaefer <schaefer(_at_)brasslantern(_dot_)com>:
It is, according to what you (and ESR) have claimed. The address from which
the queued message was sent is no longer valid. If, as you assert, it is a
MUST requirement of 2821 that the return-path be valid, then any SMTP server
that continues delivery of a queued message for which the sender has since
become invalid, is violating the protocol.
I didn't say that. There's a nuance here you're missing.
The sending MTA is conformant with RFC2821, as I read it, if it issues
a Return-Path back to the user that preserves the MAIL FROM information,
e.g. is valid at the time the message was generated. There is a MUST
attached to that requirement.
The receiving MTA is conformant if, upon ascertaining that the Return-Path
is invalid, it refuses the mail for that reason.
The fact that said Return-Path may have become invalid after transmission and
before receipt is an interesting edge case which does not affect the
meaning of RFC2821 in any way.
<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>
Asrg mailing list