is this proper behavior?

[Edward Lewis <edlewis@arin.net>]

(I appologize if this has been seen before.  I apparently forgot to 
subscribe from this address, and I suspect the mail was silently 
rejected based on looking at another account I have.)

I've been testing some delegations, trying to measure the extent of 
lameness in a few zones.  I've come across an interesting result that 
I'd like to discuss.  My hope is that this discussion is within 
charter of this list, if it is not, just tell me to stop...

It's probably clearest if I give an anonymized example of the 
behavior I am seeing.  So I'll start with an example.

According to our records, the four zones [0123].287.in-addr.arpa, are 
being hosted on ns.one.two.  When I ask ns.one.two. (substituting an 
IP address for the domain name in dig) for the SOA record of 
1.287.in-addr.arpa. I get a positive response that confirms that the 
delegation is correct.  The same happens for the 2 zone.  But when I 
query for the 3 zone, I get a server time out.  I.e., no udp response 
- no SERVFAIL, no REFUSED, no NXDOMAIN, not even a referral.  BTW, my 
queries are all non-recursive.

(This is the question that led me to ask this on DNSOP.)  Is it 
acceptable for a server to answer for some zones and not at all for 
others?  Wouldn't this cause a problem for DNS implementations that 
measure round trip times in order to select one server over another? 
I don't think such behavior is subject to a specification.

Being that I am familiar with BIND mostly, I did try to ask the 
server for its version.bind/chaos/txt, and got a FORMERR response 
back, so I have no clue as to the implementation I am testing at that 
address.  I was asking for this to see, if this was BIND, if I could 
replicate the server's behavior in order to understand what it was 
doing.

-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Edward Lewis                                          +1-703-227-9854
ARIN Research Engineer