[ On Sunday, March 16, 2003 at 19:30:38 (+0100), Florian Weimer wrote: ]
> Subject: Re: [Exim] Reaction to rude 554 greeting
>
> "Greg A. Woods" <woods@???> writes:
>
> > The problem is in people thinking they can do something about this
> > automaticaly in software without human intervention. The RFCs state
> > very clearly that human intervention is absolutely required here! Only
> > the human sender can ultimately determine the correct course of action!
>
> Unfortunately, the human sender has no way to instruct Exim to try the
> next MX, after investigating the situation (unless patching Exim
> counts as an option, of course).
You mis-understand. The human sender has all kinds of alternatives, but
which is appropriate depends entirely on the circumstances and the true
meaning of the 5xx "greeting", things no SMTP software can determine
alone.
In fact the human sender may have many ways to instruct their mailer to
deliver to an alternate destination, but again the correct choice
amongst them depends entirely on the circumstances.
> I'm all for enforcing protocol conformity. But sometimes, priorities
> are different. Not having to phone, for example. I hate to have to
> use the phone, and I'll accept some obscure misinterpretation of the
> SMTP standard if I can send mail instead. 8-)
The _ONLY_ correct priority here is to get the message back to the one
person who can ultimately decide the next appropriate course of action.
I hate to have to use the phone too, especially when the message to be
delivered does not require direct human interaction. However the
telephone is just one possiblity out of a near infinity of alternative
communications mechanisms which do not involve SMTP.
As for those which do involve SMTP, well if the sender (or at least his
or her postmaster) is savvy enough to figure out what's really wrong
here, and if your idea of trying the next MX is the best course of
action for this circumstance, then there's a very good chance that the
mail can be re-routed to that other MX host, perhaps using the ancient
fallback addressing form with a literal IP address: <user@???>
or maybe even just as <user@???>.
However it is always going to be absolutely impossible, by definition,
for the mailer software to even begin to guess at the correct course of
action without risking mis-routing the message. A message delivery
encoutering any 5xx greeting response MUST result in an immediate bounce
so that a human can decide what course of action is appropriate (and so
that at least some human will have this problem brought to their
immediate attention).
Remember the priority here is not to deliver every message to some, any,
destination, but rather to deliver messages _only_ to their obviously
intended destinations and to return all messages which cannot be
reliably routed in as quick and informative a way as possible.
Networks of SMTP systems are not like the traditional human operated
postoffice where the postmaster or any other human encoutering delivery
problems can make judgement calls based on their extensive human
experience -- SMTP systems are limited to the algorithms and data
available to them, and they must operate within the limitations required
to ensure the best possible level of interoperability between them.
--
Greg A. Woods
+1 416 218-0098; <g.a.woods@???>; <woods@???>
Planix, Inc. <woods@???>; VE3TCP; Secrets of the Weird <woods@???>
