[comp.unix.ultrix] mail/BIND/svcorder/MX annoying problem

lynch@spiff.stanford.edu (William Lynch) (05/26/90)

I can't believe we're the only ones with this problem.  We're running
3.1/2.2 on DECstations, and when using BIND, binmail really blows using MX
records for mail which should be forwarded through some remotely local site
(did I say that?).  Here's the good part though: it works fine if bind is
the FIRST thing in the /etc/svcorder file!  I suppose that binmail looks in
the svcorder file (the string is in the executable, anyway) to see if you
are BINDing, but somehow figures it doesn't need MX records unless bind is
the first thing in the file (?).  

We of course have a workaround of putting bind first in the svcorder file,
but the number of other things that breaks due to local aliasing is rather
impressive.  Does anyone have any other ideas/fixes/causes?  Can anyone
empathize?  Does anyone know if this is fixed in 4.0 (we did report it)?

Bill
lynch@umunhum.stanford.edu

PS our VAXstations don't have this problem.

avolio@decuac.dec.com (Frederick M. Avolio) (05/26/90)

In article <631@helens.Stanford.EDU> lynch@spiff.stanford.edu () writes:
>
>I can't believe we're the only ones with this problem.  We're running
>3.1/2.2 on DECstations, and when using BIND, binmail really blows using MX
>records for mail which should be forwarded through some remotely local site
>(did I say that?).  Here's the good part though: it works fine if bind is
>the FIRST thing in the /etc/svcorder file!  

I had reported this bug a while ago.  I am told by a few customers of mine 
who tested this on FT 4.0 that it is fixed, so ULTRIX 4.0 should fix this.

It was do to a programming logic error when trying to handle the service
order (which I think is a great feature).  We had local then bind listed
and where not able to send mail to hosts hat had no address record but did 
have an MX record.  Bascically, what you described we saw and I am told
it is fixed.


Fred