Re: For exim sake! Re: [Exim] Reaction to rude 554 greeting

Greg A. Woods wrote:
> [ On Tuesday, March 18, 2003 at 23:26:49 (+0100), Florian Weimer wrote: ]
>
>>Subject: Re: For exim sake! Re: [Exim] Reaction to rude 554 greeting
>>
>>Maybe the 554 greeting makes sense in the following environment: an
>>internal mail server which is used by MUAs as smarthost has been
>>migrated to a different host. The 554 message helps the users to
>>adjust their settings.
>
>
> Yes, that's plausible -- it would cause more timely response than just
> returning TCP RST would do, though it will likely still cause a lot of
> confusion and bad feelings. You'd have to be in a real bind with many
> users using a smarthost setting with a fixed literal IP# that you will
> soon lose control over (e.g. you are renumbering to a new network) to be
> forced into employing such a drastic measure.

Doesn't have to be a fixed literal address. We had (for historical
reasons, sendmail and so...) mailers which accepted and relayed mails
for our user, but which should never have been used by them, nor had
ever been documented to do so. When we moved to exim and later came to
entangle the whole mess, we found that about 20% of our users used one
of the wrong servers. Even after announcing that these would be shut
down, many continued doing so, because "it works, why change". So we
eventually stopped the mailers and replaced them by (/etc/inetd.conf)

smtp stream tcp nowait exim /usr/bin/cat cat /etc/exim/nomailserver.txt

containing "554 Please use mailout to send mail.", hoping this would
save us some support calls (I'm not sure it did). "cat" does of course
not wait for the client saying "QUIT", so it's not even in line with the
wording in the RFC.

Anyway, the only application I can devise for a 5xx on connect is to
give an explanatory message, and this only works when the sending mailer
bounces the mail. If 554 does not make the mail bounce, we would have to
have another status code for this application, 666 maybe... 8).

--
Harald Schueler