A6 lookups from clients and bind9?

[Matthew Hall]

I have numerous clients that when presented with a shortname
of a host, will attempt to do what appears to be a A6 (IPv6)
lookup:

Me on 10.0.1.116 types "ssh me at irk"

Log says:

Sep 06 14:30:20.041 queries: info: client 10.0.1.116#32843: query:
irk.foo.bar.net IN AAAA
Sep 06 14:30:20.042 queries: info: client 10.0.1.116#32843: query: irk IN
AAAA

The quadruple A is indicative of a IPv6 lookup, yes? Now, on my
bind9 servers, it seems to be trying to answer/lookup the request
from my forwarders, because there is a timeout of 10-15 seconds
or so before the client asks "the right" question and gets back
the info it needs to proceed.

Using the FQDN on the same command 'ssh me at irk.foo.bar.net' will
get an immediate response.

Now, if I switch my name server on my client from the new bind9
to the old bind8 server (we're transitioning), there is no such
delay/timeout - is there a difference in how bind8/bind9 handle
IPv6 requests - I was under the impression that bind8 didn't really
do them at all.

Linux(rh71-72) seems to have the most issues with this. My FreeBSD
box does not. Differences in the resolver libraries? truss/stracing
the running of both yields a view of a slightly different order in
the queries, hence one gets a 'better' response quicker.

Is there any way I can tell my bind9 servers to send an
immediate 'don't know/don't care' response to clients for
IPv6 queries? I think this would be the quickest way to
get things back the way they were.

Thanks for any thoughts,