> I have been seeing weird problems on NetBSD systems during the last..
> um.. 6 months or so.  Mostly, the problem shows up with failing mail
> transmission.  The messages sit in the queue and are finally returned
> to the sender with an error message.  'mailq' reports problems like
> "reply: read error" and "timeout while waiting for input".

> The weird thing is that this seems to happen _every time_ someone
> sends e-mail to a certain place.  [sending from a sunos machine
> works, using telnet works.]

When you tried telnetting, was it from the same machine that was having
trouble delivering the mail mechanically?

Someone already suggested tcp wrappers and missing or bad DNS entries
as a potential source of trouble.  But if the telnet experiment was
done from the box that was having trouble, that's not it.

The reason I'm writing this note at all is to mention that last time I
had such problems (mail repeatedly failing to a specific site) it
turned out to be that our mailer was using the default of E=\n for the
TCP mailer, instead of the mandated standard E=\r\n.  I daresay modern
installations come with the correct value set by default, but it might
be worth checking anyway.  (That same site started failing again later,
and spying on the network exonerated our machine - when we sent out the
initial SYN packet, we'd see a couple of UDP packets to ports I'd never
heard of, then a short timeout later our SYN attempt got RSTed.)

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu
		    01 EE 31 F6 BB 0C 34 36  00 F3 7C 5A C1 A0 67 1D