At Wed, 14 Oct 2015 11:55:02 +0100, Sara Dickinson <s...@sinodun.com> wrote:
> > I've read the 03 version of draft. It's very well written and I've > > not found any critical technical issues. So I basically think it's > > ready to ship. > > I have a couple of minor comments, which the authors might want to > > address/discuss before advancing the draft: > > > > 1. Section 3.2.1 > > > > DNS clients MAY include the edns-tcp-keepalive option in the first > > query sent to a server using TCP transport to signal their desire to > > keep the connection open when idle. > > > > I wonder whether we want to explicitly say something about including > > the edns-tcp-keepalive option in subsequent queries on the same TCP > > connection. > > So it may be better to explicitly say, e.g., that clients MAY > > include the option in subsequent queries > > Yes, this will definately improve the document. The following has been added > to Section 3.2.1: > > DNS clients MAY include the edns-tcp-keepalive option in the first > query sent to a server using TCP transport to signal their desire to > keep the connection open when idle. > > + DNS clients MAY include the edns-tcp-keepalive option in subsequent > + queries sent to a server using TCP transport to signal their > + continued desire to keep the connection open when idle. I have no objection to this text per se, but the described reason why the client might want to do this is different from what I intended. I was more concerned about the case where the client only includes the option in the first query and the server can't send updated notification. And, in that sense, > > We might even want to use > > a stronger requirement like "the client SHOULD occasionally include > > the option in subsequent queries if it includes the option in the > > first query”. > > I’m not sure how prescriptive the document should be on this... > > Would adding the following paragraph to Section 3.4 on TCP session management > address you concerns? [...] > + Clients might choose to send the edns-tcp-keepalive option regularly > + during a TCP session to increase the chance of being notified should > + the server modify the timeout associated with a session. The > + algorithm for choosing when to do this is out of scope of this > + document and is left up to the implementor and/or operator. while the text looks good to me, I thought it would be safer to use some RFC2119 keyword here. Combining these, what I would have proposed is something like this: DNS clients MAY include the edns-tcp-keepalive option in the first query sent to a server using TCP transport to signal their desire to keep the connection open when idle. If a client include the option in the first query, it SHOULD include the option regularly during a TCP session to increase the chance of being notified should the server modify the timeout associated with a session. The algorithm for specifically when to include the option in subsequent queries is out of scope of this document and is left up to the implementor and/or operator. -- JINMEI, Tatuya _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
