On Tue, 14 Apr 1998, Christopher Purnell wrote:
> Date: Tue, 14 Apr 1998 09:48:20 +0100 (BST)
> From: Christopher Purnell <cjp@???>
> To: Exim Users Mailing List <exim-users@???>
> Subject: Re: [EXIM] DNS
>
> Greg A. Woods wrote:
> >
> > But there *MUST* be (according to the DNS specification)! ;-)
>
> There seems to be more and more broken DNSs these days and most of
> them seem to be running NT.
Yes, and more and more bigoted sysadmins refuse to beleive their
Microsft systems are broken. Just like the ones with underscore in
hostnames.
> > I'm not surprised if Exim goes into a loop trying to find one -- I would
> > too!
> >
> > The target of an MX must be a canonical hostname, i.e. a name with at
> > least one A RR.
>
> If the target of a MX record must have an A record why look for a MX
> record if you can't find an A record?
Becuase the MX record tells what server handles mail for a domain. A
server must have an IP address.
For instance:
customerdomain1.com. IN MX 10 mailserver.bigisp.com.
customerdomain1.com. IN MX 20 backupmailserver.bigisp.com.
customerdomain2.com. IN MX 10 mailserver.bigisp.com.
customerdomain2.com. IN MX 20 backupmailserver.bigisp.com.
mailserver.bigisp.com. IN A 1.1.1.1
backupmailserver.bigisp.com. IN A 1.1.2.2
If the mailservers didnt have IP's, how can you deliver mail for the
two customer domains to it?
It would be wrong to say:
customerdomain1.com. IN A 1.1.1.1
Becuase, for one, you dont get the backup handling that MX provides,
and two, and that doesnt reflect reality - the host at 1.1.1.1's name
is mailserver.bigisp.com, not customerdomain1.com. I wont expound on
my personal abhorance to assiging an IP directly to a second-level
domain.
> > Exim should probably bounce the message with a "host not found" error,
> > perhaps with some detail saying that the MX is broken. I suppose it
> > could also freeze the message and alert the postmaster so she an go and
> > bug the owner of the MX for a fix.
>
> No, the correct behaviour is to deliver to one of the other targets of
> the MX record.
That would really only be the correct behavior if the MX record was
correct but the host was unreachable. The correct behaviour would be
for the DNS admin to set records for their domain that reflect some
resemblence to both RFC and to sanity.
> This is really causing us problems and I really need some sort of fix
> for it. Especially since some idiot moved all our machines from
> Oxford, where we are, to London.
So telnet to them and fix the problem. Or are these NT machines? Thats
one of the primary reasons NOT to use NT machines for critical Internet
servers - remote administration is at best complicated and at worst
impossible.
>
> --
> Christopher John Purnell
>
> --
> *** Exim information can be found at http://www.exim.org/ ***
>
>
--
*** Exim information can be found at
http://www.exim.org/ ***
