Bind 9 truncating responses

[stephen.vickers at n3k.co.uk]

Can anybody comment?

First things first: I understand that 50 NS records in a zone are not good practice however situations require that this be the case.

I have been completing some testing where I have approx 50 NS records within a zone.  When I place this zone on a Bind 9 Server and query for an A record in this zone from nslookup Bind places the answer in the answer section correctly.  It then jams as many NS and Glue records in the additional and authority section as possible.  As not all of the NS and Glue records fitted in the packet the Bind 9 Server sets the TC bit in the DNS header.  This same test on Bind 8 shows different behaviour as Bind 8 does not set the TC bit.

Should Bind 9 be doing this as it supplied the answer?  Setting the TC would cause most implementations to retry the query via TCP causing more work for the Server and using up a TCP resource.

Are the NS and A records considered required/important for the response to an A record query?

Cheers