> On 4 Jan 2016, at 18:18, Brian Haberman <br...@innovationslab.net> wrote: > > ---------------------------------------------------------------------- > DISCUSS: > ---------------------------------------------------------------------- > > I support the publication of this document, but I have a point I want to > discuss to help with the clarity of the spec. > > Section 3.2.1 says that clients send this option with the first query > sent on a TCP connection and Section 3.2.2 says it should honor the > timeout provided by the server and close the socket when appropriate. > What is not discussed is how the client should manage the timer with > respect to the reception of multiple query responses that may, or may > not, include edns-tcp-keepalive option. Section 3.3.2 says the server MAY > send the option, so it is up to the server to decide when to include the > option and the corresponding timeout value. Should the client's timer > simply reflect the value sent in the latest response? The smallest > remaining time? > > I think a few sentences on client timer management would be beneficial.
Hi, Thanks for the feedback. The intention in section 3.2.2 was that the client should update the timeout whenever receiving a new value in any response. Would it be enough to clarify by changing the last sentence in the second paragraph to: “It SHOULD honour the timeout received in that response (overriding any previous timeout) and initiate close of the connection before the timeout expires.” Regards Sara. _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
