puzzled by NS records from referrers

Marc_Lampo at hotmail.com Marc_Lampo at hotmail.com
Mon Feb 7 22:29:26 UTC 2000


Hello (again),

I'm puzzled by the NS records returned by an authoritative NS for a tld
that also happens to be slave for the domain in question.

This is the situation :
ns.domain.tld. is the authoritative for tld.

In the zone file there is :
customer.tld.  NS  ns.domain.tld.
 (observe : this tld. server is itself slave for customer.tld)
customer.tld.  NS  ns.customer.tld.
 (+ required glue record)

The master for customer.tld. has :
customer.tld.  NS  ns.domain.tld.
customer.tld.  NS  ns.customer.tld.
customer.tld.  NS  ns.other-domain.tld.
 (observe : the master announces a superset of what the tld. server
knows)

(I don't really want to start a discussion about the pro/contra of
having the tld. server being slave for subdomains within tld. as well;
I'm convinced it shouldn't be.  Just trying to understand Bind8's
behavior)

In this given situation ns.domain.tld. learns NS-RR's from two sources :
1. the zone file of tld. (holding 2 NS's)
2. data from the master obtained via zone transfert (holding 3 NS's)

I actually expected this name server to return the 3 NS's when asked
for, but it turns out it sticks to the 2 in the tld. zone file.
So it looks like it ignores (at least the NS's) that come in from the
master.  But I fail to see why ?

Also noteworthy is that slave servers of the tld. domain (that are not
slave for customer.tld.) do return various answers : some return 2 NS's,
others return 3 NS's. (which seems unrelated to the version of the
server or the "recursion available" being set or not).

I'm puzzled.  Do you have enough info to throw a light on this ?

Many thanks,

Marc


Sent via Deja.com http://www.deja.com/
Before you buy.



More information about the bind-users mailing list