problem with ixfr

[Kevin Darcy]

Carl Byington wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> ns.five-ten-sg.com contains the zone blackholes.five-ten-sg.com which is
> fairly large (137K records). ns1.five-ten-sg.com used to do IXFR
> transfers, but after moving that machine to a different network, it now
> only does AXFR transfers.
>
> from ns1, we can 'dig blackholes.five-ten-sg.com ixfr=2006052704 @ns' and
> get the delta update.
>
> from ns, we can 'dig blackholes.five-ten-sg.com ixfr=2006052704 @ns1' and
> get the delta update.
>
> Do those dig commands not emulate what bind itself is doing internally to
> request the zone transfer?
>
>   
It should look somewhat similar. The only potential difference that 
comes to mind immediately is the use/non-use of EDNS0 and/or the EDNS0 
buffer size that is negotiated between the master and slave. Potentially 
that might result in a different packet size, one that might be more 
likely to be dropped/truncated/corrupted by intermediary firewalls, 
routers or other network devices. You could try turning off EDNS0 
completely in respective "server" statements, to see if that has any 
effect on the behavior.

                                                                         
                           - Kevin