On Dec 1, 5:09am, Eric S. Raymond wrote:
}
} Bart Schaefer <schaefer(_at_)brasslantern(_dot_)com>:
} > } RFC2821 says the Return-Path has to be the same address that was
} > } presented to the MAIL command.
} >
} > So what? MX1 isn't any party to the protocol exchange between MTA1 and
} > MX2. The reverse-path is completely valid as far as MTA1 can tell; if
} > MX2 responds 5xx, MTA1 will do the right thing. You still haven't shown
} > me how or why any of this affects what happens if MX1 becomes party to
} > a separate SMTP transaction initiated by MX2, whether involving the same
} > or any other mailbox.
}
} I don't see how the presence of MXes matters.
Hector's caller-ID proposal determines the validity of an address coming
from MTA1 by initiating a new SMTP transaction with the corresponding MX,
in this case MX1. We wouldn't be having this discussion at all if the
presence of MX1 were not required.
} Are you maintaining that what's in the MAIL FROM need not be a valid
} reply address?
I am maintaining that a separate SMTP transaction among MX2 and MX1 is
not a reliable test of whether what's in MTA1's MAIL FROM is a valid
reply address, because the existing protocol definition cannot require
that it be a reliable test.
If that means the MAIL FROM is not a valid reply address, then yes, I
am asserting that nothing I've yet seen quoted from RFC2821 requires
that what's in the MAIL FROM need be a valid address to which MX2 can
reply. I further assert that if 2821 did require that, it would then
prohibit those edge cases that you've previously dismissed; which edge
cases it manifestly does not prohibit.
_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg