[dns-operations] Missing .us and GTLD records??

[Florian Weimer]

* bert hubert:

> On Sat, May 22, 2010 at 03:54:49PM +0000, Florian Weimer wrote:
>> * bert hubert:
>> 
>> > It might be a grand idea to automatically switch on IPv6 for outgoing
>> > queries if root-priming succeeds over V6.
>> >
>> > And perhaps switch IPv6 off again if none of the roots is reachable over
>> > IPv6.
>> 
>> I think the question is: why do you want to use IPv6 at all if IPv4 is
>> available?
>
> "it might be faster" - when determining which nameserver address + family to
> talk to, everybody does so on the (banded) historical RTT. The IPv6 address
> might simply be faster.
>
> Unlikely as this seems, I've seen an example of this,

I've seen that, too.  Two ISPs, transit customers of the same ISP,
route traffic away from that upstream, probably for cost reasons, but
they use their upstream's IPv6 tunnel broker and have insufficient
IPv6 connectivity to route away from that, too (or don't bother about
the traffic levels).  So IPv6 has actually lower latency.

But in the majority of cases, it's going to be way slower, and often
you wait for a response that will never arrive.  If you can detect
better response times out without exposing latency increases, all the
better, but I think it's going to be tough.  (Unbound could do it,
with its background requery of popular names prior TTL expiration.)

-- 
Florian Weimer                <fweimer at bfk.de>
BFK edv-consulting GmbH       http://www.bfk.de/
Kriegsstraße 100              tel: +49-721-96201-1
D-76133 Karlsruhe             fax: +49-721-96201-99