Re: [dnsext] RFC3596: A/AAAA records and additional section processing

[Jehan Pagès <[e-mail]>]

Hi,

2011/3/4 Colm MacCárthaigh <[e-mail]>:
> The good news is that it is standard - section 3 of the same RFC says;
>
> "3. Modifications to existing query types
>
>   All existing query types that perform type A additional section
>   processing, i.e., name server (NS), location of services (SRV) and
>   mail exchange (MX) query types, must be redefined to perform both
>   type A and type AAAA additional section processing.  These
>   definitions mean that a name server must add any relevant IPv4
>   addresses and any relevant IPv6 addresses available locally to the
>   additional section of a response when processing any one of the above
>   queries."
>
> (e.g. dig MX example.stdlib.net @ns-191.awsdns-23.com.) .
>
> The section you've quoted is just pointing out that a query for a
> q-tuple of type AAAA would not normally by itself trigger any
> additional-section processing. As worded, I think it's actually wrong.
> A query of type AAAA might trigger additional section processing if
> the query arrives at the parent of a delegated zone. For example the
> query;

I did read this section you quote but from my understanding this is a
different case. It says that if some query already makes *A additional
section processing*, then it should do both A and AAAA additional
section processing. Hence the examples: NS, SRV, MX which all do A
additional section processing.

This is not what I am talking about. I am talking about *direct* A
query (answer section) and *direct* AAAA query. And I believe the
section I quoted is the one answering how it is currently
standardized: no additional section processing.

> dig AAAA www.google.com @a.root-servers.net.
>
> plainly does trigger additional section processing. (though one might
> argue it was the name, not the type, that pulled the trigger).

That's what I say. Some servers out there already do this because that
indeed seems so logical. And I agree with this behavior. Simply this
is not the standardized recommended behavior. This is why I ask why,
and maybe we could make it the default standardized behavior unless
someone comes up with good reasons against.

Jehan

> 2011/3/3 Jehan Pagès <[e-mail]>:
>> Hi all,
>>
>> I have a small question about the IPv6 resource records extension (RFC3596).
>> As section 2.3 states:
>> « A type AAAA query does not trigger additional section processing. ». Fine.
>>
>> But I noticed that some DNS servers (like the one installed on my
>> server, a bind 9.7.2) would return the A records in additional section
>> on a AAAA query and AAAA records in additional on A queries.
>> And when I thought about this, I realized it was a pretty good logics.
>> Usually when people will try to contact a server, they will want both
>> A and AAAA records.
>> So the question is: why wouldn't this be standard?
>> Thanks.
>>
>> Jehan
>> _______________________________________________
>> dnsext mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/dnsext
>>
>
>
>
> --
> Colm
>