Re: [Exim] Using an external smtp server

| >Where a Novell Netmail server is trying to send mail to some domain, and
| >encounters a 5xx at RCPT time from the primary MX, it (netmail) merrily
| >proceeds to try the exact same delivery on the backup MX(es).
|
| How bloody horribly broken. And here I thought NetMail was a decent MTA (I
| havent played with it yet, but from what I can see of their myrealbox.com
| demo vehicle / freemail account, it looks pretty decent, does AUTH, TLS and
| all)

Well, I got a myrealbox account and sent a test message to an unknown user
back at home. Sat back and watched the logs:

Rejection by our primary MX:

2003-06-02 12:39:52 H=smtp-send.myrealbox.com [192.108.102.143]
F=<chris-e@???> rejected RCPT <qwerty@???>:
unknown local-part "qwerty" in domain "mxtest.gla.ac.uk"

The a second later they try our backup MX:

2003-06-02 12:39:53 H=smtp-send.myrealbox.com [192.108.102.143]
F=<chris-e@???> rejected RCPT <qwerty@???>:
unknown local-part "qwerty" in domain "mxtest.gla.ac.uk"

A colleague (hi alan) writes:

-----------------------------------------------------

" RFC2821 says

5yz Permanent Negative Completion reply

The command was not accepted and the requested action did not
occur. The SMTP client is discouraged from repeating the exact
request (in the same sequence).

I suppose they _could_ argue that by trying the secondary MX they aren't
repeating the "exact request", but I wouldn't be happy with that
interpretation.

[...]

The message of the subsequent text in 2821 seems to me to say that a
retry is only appropriate after human intervention to correct whatever
issue the 5xx was reporting. But it seems not to be an absolute. "

-----------------------------------------------------

I'm itching to tell Novell they're wrong (we're customers) but it seems
not so clear :-(

Chris

--
Chris Edwards, Glasgow University Computing Service