Re: portables, mail, and sporadic connectivity
Craig Metz writes ("Re: portables, mail, and sporadic connectivity "):
> I had a setup where diald was being used for a SLIP link. There
> would always be a route for network access to the outside world, but the
> link would go dead sometimes and require diald to sink packets while it
> tried to bring the line back up. This caused DNS requests to fail (server
> failure message, not the no entry message) because my name server couldn't
> talk to the root nameservers. When Smail received this failure indication,
> it bounced the message back to sender instead of trying to retry the lookup
> and/or queue for later processing tries. This apparently has something to
> do with the new queue/DNS interaction in 3.1.29; 3.1.28 didn't have this
> particular problem (but it had others). Smail also doesn't like having the
> rug pulled out from under it when it's talking SMTP, though it usually
> recovers.
If Smail is doing this it is a bug. Can you reproduce the problem
with Debian's Smail package ? If so please let me know and I shall
investigate and fix the problem.
Temporary DNS lookup failures (eg, servers being unavailable) should
be treated as soft errors, leaving the messages in the queue; only an
authoritative negative response (or a prolonged period of
unavailability) should cause messages to be bounced.
> Granted, the root of my problem here was lousy connectivity, but it
> managed to get Smail confused. When I installed Sendmail 8.6.12, these mail
> problems went away. If only there were a chance that Sendmail could be
> made secure, I'd feel better about using it. For something like Debian,
> however, Smail is probably the winner because (a) Debian prefers GPL software
> (b) Smail works fine for 99.9% of the cases.
I think that the root of your problems was probably a mistake in your
Smail build's compile-time configuration. If it was your Smail build
then it was probably your mistake; if it is my Smail build I'd be very
interested to hear more.
Ian.
Reply to: