Speaking for myself, I love Evan's idea with picking one rrset. It is
simple. Would the simplest be to grab the first rrset it finds?
tim
On 3/25/15 5:28 PM, Matthijs Mekking wrote:
All,
First of all, I really like Evan's idea. It's nice and simple to implement.
Also, I think we can just pick any one RRset. In amplification it is no
worse than querying for the corresponding QNAME/QTYPE. And adding search
behaviour or learning capabilities adds unnecessary complexity that
costs additional processing.
Best regards,
Matthijs
On 03/25/2015 05:11 PM, Darcy Kevin (FCA) wrote:
Intriguing. I think there might be a temptation to take this idea one
step further and say that preference should be given to the
*smallest* RRset possible -- to reduce the degree of amplification.
But that would be a mistake. It would, for instance, discriminate
IPv4 over IPv6 (since A records are smaller than AAAA records) and
certain services over others (SRV records are larger than MX records
are larger than A records). When the client doesn't get the RRset it
wants from the QTYPE=* query, it'll follow up with one or more
subsequent queries for the type(s) it cares about. Not very
efficient, and the end result is more "amplification" than was hoped
for.
Ideally, the QTYPE=* responder should *learn* what is most *useful*
to the client, e.g. if the client usually follows up a QTYPE=* query
with a QTYPE=MX query, whenever it doesn't get MX as the chosen QTYPE
of the response, then try picking MX as a response to the QTYPE=*
more often and see if the client is happier, i.e. the number of
followup queries goes down.
Alas, this level of adaptive behavior would require holding more
per-client state information in the server than is comfortable, and
in any case, is probably beyond the algorithmic capabilities of most
DNS implementers...
- Kevin
-----Original Message----- From: DNSOP
[mailto:dnsop-boun...@ietf.org] On Behalf Of Tony Finch Sent:
Wednesday, March 25, 2015 5:57 PM To: Evan Hunt Cc: dnsop@ietf.org
Subject: Re: [DNSOP] another way to minimize ANY responses
On 25 Mar 2015, at 20:50, Evan Hunt <e...@isc.org> wrote:
Last night the dumb-idea fairy visited me as I was falling asleep,
and suggested that another way to reduce the impact of ANY queries
would be to pick *one* rrset and return just that.
Very nice :-)
Tony. -- f.anthony.n.finch <d...@dotat.at> http://dotat.at
_______________________________________________ DNSOP mailing list
DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
_______________________________________________ DNSOP mailing list
DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
_______________________________________________
DNSOP mailing list
DNSOP@ietf.org
https://www.ietf.org/mailman/listinfo/dnsop
_______________________________________________
DNSOP mailing list
DNSOP@ietf.org
https://www.ietf.org/mailman/listinfo/dnsop
