How many NS records should be in the 0.0.127.in-addr.arpa zone?

[D. Stussy]

On 21 Sep 1999, Bill Larson wrote:
> > > > According to the DNS and BIND book the zone file for 0.0.127.in-addr.arpa
> > > > should contain an NS record for each server in your domain (see the 3rd
> > > > edition page 67).  First question - is this a correct assumption?
> > > 
> > > This doesn't seem correct.  I don't have my copy of DNS & BIND handy
> > > so I can't check your reference.
> > 
> > D&B does NOT say that.  It gives an example, in which there are two (2)
> > NS records instead of 1.  It makes no statement why it does so, or
> > whether you should use 1 or 2 or N, N > 2.
> > 
> > I realized today why that might be.  Some systems start other network
> > aware programs before 'named'!  ;-?  If they cannot reach their own
> > local 'named', they might be able to reach the other one.
> > 
> > Maybe I'll change mine to point to two servers, now.  ;-}
> 
> I'm confused as to why having more than one NS record for the
> 0.0.127.in-addr.arpa zone would be useful.
> 
> If every name server is a master for this 0.0.127.in-addr.arpa zone,
> then a query of this zone send to ANY server would always return an
> authoritative answer.  A server can ONLY respond authoritatively to
> a query if the server is identified with an NS record as being
> authoritative for the zone.
> 
> Now, as to the point of the system starting network applications prior
> to starting named, if named isn't running then named cannot return ANY
> answer - including identifying a second name server to query for a
> zone.  This is an issue that can only be resolved by the use of
> resolv.conf identifying more than one name server and having a
> non-local named supply this DNS information, or possibly using an
> nsswitch.conf type file directing the gethostby* library routines to
> use /etc/hosts, DNS, and/or NIS for resolving DNS type queries.

However, "named" can respond to those other applications if the response is
coming from named on a DIFFERENT HOST.....  I'm not certain what you mean by
"non-local" (not same machine, not same network, not same geography, etc.?).

> Again, for what purpose would having more than one NS record be for the
> 0.0.127.in-addr.arpa zone?  I will hold to identifying ONLY a single
> server for this zone.  This would also avoid the problem with NOTIFY
> packets for the zone which started the whole thread (as identified by
> Jim Reid's posting).