NB: internic.net appears to be looping at the moment :-((( e.g.
The original message was received at Thu, 12 Sep 1996 02:45:28 -0400 (EDT)
from uucp@localhost
----- The following addresses have delivery notifications -----
<markk@???> (unrecoverable error)
----- Transcript of session follows -----
554 Too many hops 19 (17 max): from <piete.brooks@???> via localhost,
to <markk@???>
> > Neal> If I try to send mail to a domain which has 2 MX records, and the
> > Neal> lower numbered MX returns "connection refused", should the 2nd MX be
> > Neal> tried? I'm not sure if this is what is happening.
> > Yes. This is the point of MX records. In fact, here at the internic,
> > the mail scheme is set up so the lower MX value *always* gets a
> > "connection refused", so if this scheme didn't work, I would never get
> > any mail :-)
> I think I'd be hearing loud screams from many Exim users if it were not
> trying alternative MX records. :-)
Trying alternative MX RRs is not what he asked about -- it was specifically
when having received "connection refused" that was the concern ...
> David: It seems a bit strange to cause every mailer in the world that
> mails you to try a connection that is always going to fail.
I was going to use such a scheme in the early days of JIPS (ShoeString as it
then was ...) -- it's a neat solution, but unfortunately the penalty is paid
in the normal case, rather than the exception ...
The basic idea is how (without playing with DNS servers in nasty ways which
typically work, but will break occassionally) to get external email to use a
relay, while internal email goes direct ....
> There will be some waste of machine time and net resources.
A bit -- but not *that* much.
[[ In the end we decided it was too expensive ]]
> Is it really the only way?
It has a certain elegance :-)
