IPv6 and IPv4 resolver preferences

Felipe Alfaro Solana felipe_alfaro at linuxmail.org
Fri Sep 17 09:46:49 UTC 2004


On Sep 17, 2004, at 02:11, Ulrich Drepper wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Felipe Alfaro Solana wrote:
>
>> from hostA to nameserver, standard query AAAA hostB
>> from nameserver to hostA, standard query response AAAA 2000::1
>> from hostA to nameserver, standard query A hostB
>> from nameserver to hostA, standard query response A 192.168.0.1
>> from hostA to nameserver, standard query A hostB
>> from nameserver to hostA, standard query response A 192.168.0.1
>> from hostA to nameserver, standard query AAAA hostB
>> from nameserver to hostA, standard query response AAAA 2000::1
>> from hostA to hostB(192.168.0.1), 32805 > telnet [SYN] Seq=0 Ack=0
>> Win=5840...
>>
>> As you can see, the hostA resolver is querying twice for A and AAAA 
>> RR,
>> which seems completely strange.
>
> It's not strange at all.  The second set of lookups is for the 
> canonical
> host name.

I think I don't understand you.

When running "telnet computer1" and looking at the packets captured by 
Ethereal, I can see _exactly_ four queries, all of them asking for the 
RR host record of "computer1.mydomain.com", that is, four queries 
asking for the same FQDN. Of the four queries, the first one asks for 
the AAAA RR, while the remaining three ask for the A RR. These three AA 
RR queries are all identical (except in the Transaction ID field).

Thus, I don't understand why the resolver is simply asking three times 
for the A RR. Even if the resolver does not make any difference between 
FQDNs and non-FQDNs, I think this can be optimized a little, and 
simplify it to just, at most, two queries.

Isn't this reasonable?

>
> Anyway, all this is mood with the current glibc code.  Once you get
> 2.3.3-53 or higher, try it again.
>
>
> And IPv6 addresses are always looked but if requested and there is an
> IPv6 interface defined on your system.  If you don't want any IPv6
> lookups, remove all IPv6 interfaces.

In my case, it's just the other way around. I DO want IPv6 queries, and 
I DO want IPv6 to always have precedence over IPv4. Anyways, I'll get 
glibc-2.3.3-53 as soon as it's available.
Thanks!





More information about the fedora-devel-list mailing list