Re: Fwd: New Version Notification for draft-wouters-edns-tcp-keepalive-00.txt (fwd)
Tony Finch <[email protected]>
| Newsgroups | gmane.ietf.dnsext |
|---|---|
| Message-ID | <[email protected]> |
Joe Abley <[email protected]> wrote: > > So the key observation here, if I'm reading it correctly, is that > clients might reasonably try to re-use an open socket even if the > edns-tcp-keepalive option is not present, anticipating a short (but not > impossible) timeout on the server side. Right. There are (basically) three kinds of server: Working edns-tcp-keepalive servers support persistent TCP and advertise this fact. RFC 5966 clients and edns-tcp-keepalive clients will both use persistent TCP connections and everyone is happy. Working RFC 5966 servers support persistent TCP but do not advertise this. RFC 5966 clients will use persistent TCP connections, but edns-tcp-keepalive clients will think the server is in overload and will not use persistent TCP even though they could. Sadface. Broken servers may or may not advertise support for persistent TCP. An edns-tcp-keepalive server may be broken because there is a nasty middlebox in the way, or perhaps it got overloaded without being able to signal this to the client. So clients have to detect broken servers based on their ability to keep a TCP connection open to the server, and in any case they have to be able to recover gracefully from an abruptly closed connection. I'm a bit doubtful that there is any benefit in an explicit "I'm broken" signal from the server using the edns-tcp-keepalive option. I like the aims of the draft and I think it is useful for the server to be able to say to the client, please use TCP, and please close idle connections within N seconds. But the current version is a wee bit buggy. Tony. -- f.anthony.n.finch <[email protected]> http://dotat.at/ Forties, Cromarty: East, veering southeast, 4 or 5, occasionally 6 at first. Rough, becoming slight or moderate. Showers, rain at first. Moderate or good, occasionally poor at first. _______________________________________________ dnsext mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsext