Dave Crocker wrote:
Are you guys saying that current implementations do not conform
to the spec or are you saying that the spec should be changed to
tell people to do things differently?
Some interpret the spec. as saying "I can always accept mail and
bounce it later if I don't like it". Others intrepret is as "I
can't accept mail unless I'm almost sure that I can (and will)
deliver it, or if I'm almost sure that a later bounce (NDR) won't
hit innocent bystanders."
Another group misinterprets it as "dropping suspicious mails is
better than bouncing them". That's wrong if they're not _very_
sure that the mail is bad.
The other two interpretations are covered by "the originator as
indicated in the reverse path". The first group simply believes
that the reverse path is okay. The 2nd group takes into account
that this "indication" can be forged. As it's the case in most
mails today.
The first group claims that they MUST send NDRs to follow the
spec., the second group says that receivers SHOULD reject mails
at their border MX to follow the spec.
The latter isn't explicitly stated in 2821. It's a consequence
of an interpretation of "originator as indicated in the reverse
path", plus the observation that this indication is generally
bogus at (or behind) an MX.
draft-hutzler-spamops tried to address this issue, but only for
MDAs, typically behind the MX when it's too late.
Frank