strange slowness?

[Jim Reid]

>>>>> "Clifton" == Clifton T Sharp <clifto at clifto.com> writes:

    Clifton> I've been blaming the occasional slow response of 8.2.2p5
    Clifton> on busy servers or slow paths to them. But I just saw
    Clifton> something that makes me wonder if I don't have a problem.

    Clifton> I tried to reach web.utk.edu and got a DNS failure. So I
    Clifton> hit a shell and did `dig web.utk.edu`; timeout. `dig
    Clifton> utk.edu`; timeout. `dig utk.edu` again; timeout again.

    Clifton> `dig edu.` returned a.root-servers.net for SOA in 148
    Clifton> ms. Then `dig @a.root-servers.net utk.edu.` returned NS
    Clifton> records for utk.edu in 95 ms.

    Clifton> Immediately afterward,`dig utk.edu.` timed out.

    Clifton> Now, didn't my two-step lookup exactly duplicate what
    Clifton> BIND is supposed to do to look up an address?

No. Your experiments with did didn't do a "dig utk.edu ns" to get the
NS (and maybe their A records) for the utk.edu domain. Then you didn't
resolve any of those NS records that didn't get A records in the
Additional Section of the reply for the utk.edu lookup. Then you
didn't do:
	foreach i in A-records-for-utk.edu-NS-records
		if (dig @$i www.utk.edu )
			break
	end

That's what your name server was doing when it was trying to resolve
www.utk.edu. And if your name server was doing forwarding, even more
time could be spent trying to lookup these names.

So it looks as if your name server was unable to contact any of the
name servers for utk.edu, hence the lookup timeouts.