Ray Bellis <[email protected]> wrote:
> On 14/10/2015 13:53, Sara Dickinson wrote:
> > Ah, OK. So, would the following be acceptable?
> >
> > Since pipelined responses over TCP can arrive out-of-order, clients MUST
> > match
> > responses to outstanding queries using the DNS query ID and the
> > transport tuple (protocol, source and destination addresses, source and
> > destination ports). If the response contains a question section the
> > client MUST
> > match the QNAME, QCLASS and QTYPE fields.
>
> Strictly speaking, the O/S should be taking care of matching the
> transport tuple before it ever reaches the client, and that's already
> called out in the last sentence of ยง5 where it's required that the
> response must arrive on the same TCP connection as was used for the request.
Yes. Perhaps "transport tuple" is an excessively clever way of saying
"same connection". When I wrote that I wanted to allow independent choice
of query ID per connection (rather than implying a single global ID).
Also, for UDP if you don't match the transport tuple then you lose (some
of) the benefit of port randomization, I think (depending on how many
ports you keep open).
Tony.
--
f.anthony.n.finch <[email protected]> http://dotat.at/
Fair Isle: Southwest 5 to 7. Moderate or rough, occasionally very rough at
first in west. Fair. Good.
_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop
