Re: [Exim] Failing behviour based on SMTP codes.

On Wed, 18 Oct 2000, Greg A. Woods wrote:

> [ On Tuesday, October 17, 2000 at 21:27:46 (-0700), Trevor Sky Garside wrote: ]
> > Subject: Re: [Exim] Failing behviour based on SMTP codes.
> >
> > My thoughts on this would be that one should treat a 5xx response from a
> > *primary* (i.e. lowest preference number) MX as permanent, but if the
> > response came from one of the less preferable MX hosts, it should be treated
> > as if that host was unreachable. This would be useful in this example:
> >
> > MX 5 mail.domain.com. (down)
> > MX 10 backup.otherdomain.com. (temporarily broken)
> > MX 15 never.broken.mail.host. (working great)

I had this thought also, but concluded that you will probably just
generate spurious retries in the normal case. If MX 10 is correctly
giving 5xx messages, you don't want to bother with MX 15 which should do
the same. A well configured temporarily broken MX 10 will give 4xx
responses, so we're only saving people with badly behaved systems. Not
worth it.

> There's almost no valid reasons
> remaining in this modern world to ever use secondary MXers anyway, and
> indeed many good reasons other than just this to avoid them altogether.

I have sympathy with this viewpoint, but I don't think it is as true as
you might think. There are yet still excellent reasons for having backup
MXes.

A common one is for a dial-on-demand or ISDN type corporate connection to
an upstream ISP, which is the case I'm interested in. The primary MX is
intermittently connected. The backup MX is always available.

There are also still parts of the world where connectivity remains
intransitive (A can see B and B can see C but A cannot see C) where backup
MXes can help.

John