[10077] in Athena Bugs
missing NS records in query response
daemon@ATHENA.MIT.EDU (Marc Horowitz)
Tue Nov 17 03:18:19 1992
To: bug-hesiod@Athena.MIT.EDU
Cc: bugs@Athena.MIT.EDU
Reply-To: Marc Horowitz <marc@MIT.EDU>
Date: Tue, 17 Nov 92 03:18:05 EST
From: Marc Horowitz <marc@Athena.MIT.EDU>
If I query my local named for marc.grplist.ns.athena.mit.edu, type
TXT, class HS, the response which comes back claims to have one
question record (correct), one answer record (correct, long, so it's
fragmented into two [length,data] tuples), and three nameserver
records. All of the nameserver are missing from the reply packet.
If I make the same request via a VC connection (TCP), then the packet
has the same header, and includes the nameserver records. It even
includes the A records of the nameservers as additional records.
If I make a request for probe.grplist... via UDP, it fails like mine
does. Via TCP, it works fine, and includes the A records. probe's
records needs fragmentation.
If I make a request for eichin.grplist... via UDP, it works,
specifying 3 nameserver records, and returning them in the packet.
Via TCP, it works, and includes the A records. eichin's record needs
fragmentation.
If I make a request for sipb0.grplist... via UDP or TCP, all the
nameserver and additional A records are there. sipb0's record doesn't
need fragmentation.
My guess is that for UDP, the code is attempting to remove nameserver
and/or additional records to keep the packet size down, but is failing
to update the header if it has to remove nameserver records from the
response. I tried looking for the problem, but the source is rather,
aaahhh, opaque. I decided against tracing goto's :-)
If you want me to do any more testing, instrumenting the perl code for
debugging is really easy. There's a lot of debugging info in there
already. Let me know.
Marc
