Re: Fwd: New Version Notification for draft-wouters-edns-tcp-chain-query-00.txt
Paul Wouters <[email protected]>
| Newsgroups | gmane.ietf.dnsext |
|---|---|
| Message-ID | <[email protected]> |
On Wed, 11 Sep 2013, Dave Lawrence wrote:
> Joe Abley writes:
>> Isn't "keep the TCP connection open" superfluous, since in DNS the
>> responsibility for closing the socket lies with the client, not the
>> server?
>
> Yes, and no. This is definitely what the spec says,. For years many
> operators have believed it to be a misfeature of the original spec and
> have subverted it in various ways, including short timeouts before the
> server kills the connection if a client EOF is not forthcoming.
>
> It would be useful to have a way to negotiate the TCP connection
> behavior, including options like signalling that the server may close
> the connection as soon as the response is sent, or that the client
> wants to send more than one request. I don't think this meaning
> should be bundled into the DNSSEC chain query option, however.
I've heard several people tell me this now and I agree. The question is:
1) do we want to clarify in a separate draft that the TCP connection
can remain open for re-use?
2) do we want to use another edns0 option to ask the server to keep the
TCP connection, which the server indicates by returning this option
in its TCP reply?
It seems to be 2) is probably a better idea, and I'm happy to write that
up in a separate document.
Paul
_______________________________________________
dnsext mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsext