> On 10 Oct 2015, at 17:34, Stephane Bortzmeyer <[email protected]> wrote:
> 

Joe, Stephane, (and I just saw Tim too)

Many thanks for the detailed reviews.

> 
> In the mean time, the issue I see is in section 7 "Since pipelined
> responses can arrive out-of-order, clients MUST match responses to
> outstanding queries using the ID field and port number." This has been
> recently discussed in the DPRIVE working group and seems questionable,
> specially for TCP (since the source and destination port are fixed for
> a given connection). Clients should use the ID field and
> QCLASS+QTYPE+QNAME instead, to demultiplex.


On this point, I agree that the wording could be clarified by talking about 
matching responses on a single TCP connection, rather than using port number.

Section 6.2.1. of the draft says:

"When sending multiple queries over a TCP connection clients MUST take
 care to avoid Message ID collisions.  In other words, they MUST not
 re-use the DNS Message ID of an in-flight query.”

So under normal circumstances matching on just the Message ID should be 
sufficient for TCP, which was the reason the QCLASS+QNAME+QTYPE
part was removed when changing the ‘must' here to ‘MUST' in the last revision 
of this draft.

However it is true that requiring matching on all the fields would be more 
consistent with Section 9 of RFC5452 (even though that document is mostly UDP 
focussed) and there seems to be consensus to do that. 

Regards

Sara. 
_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop

Reply via email to