help needed solving DNS puzzle (possible referral loop)

[Barry Margolin]

In article <01a401bee501$42c223e0$b477a8ce at acmebw.com>,
Cricket Liu <cricket at acmebw.com> wrote:
>Barry Margolin <barmar at bbnplanet.com> wrote in message
>news:<kUFs3.69$m84.1007 at burlma1-snr2>...
>> It's not a Microsoft bug (wow, I never thought I'd be saying that).  If a
>> domain is delegated to a server, and that server doesn't have the domain
>> installed on it, DNS lookups in that domain should be expected to fail
>> sometimes.  The querying DNS server is not supposed to keep trying servers
>> until it gets one that gives an answer that it likes.  The only time you
>> try an alternate server is when the server you tried doesn't respond at
>> all.
>
>Actually, I don't believe that's true of BIND.  BIND has an internal
>mechanism to keep the name server from querying a lame server for what's now
>known at the lame TTL, by default ten minutes.  See
>http://www.isc.org/bind8.2/options.html and nslookup() in ns_forw.c.

This is not a requirement in the DNS protocol, though.  It's nice that BIND
tries a little harder than required to avoid bad servers, but I wouldn't
claim that Microsoft's behavior is a bug.  I'm hardly a Microsoft
apologist, but I prefer to blame them when they're actually violating the
protocol, which they do enough.  A lame delegation is a bug in the
delegator's or delegatee's configuration, and other DNS servers are not
required to expend lots of effort trying to work around them.

-- 
Barry Margolin, barmar at bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.