[Standards] s2s and gracelessly broken streams
dave at cridland.net
Wed Apr 4 15:07:44 UTC 2007
On Wed Apr 4 15:48:30 2007, Richard Dobson wrote:
>> I would also say "don't impose a requirement to send
>> non-information bearing octets around purely to maintain a
>> connection", too. Basically, if you say "s2s connections MAY be
>> timed out when idle, and such a timeout MUST NOT be lower than 30
>> minutes", then this gets translated into implementations as
>> "servers MUST send their peers some useless stanza every 25
>> minutes", which - supposing we take Tony's figures - means that he
>> might be getting or sending 3.3 of these every second on average.
> As a note though we must ensure we don't specify particular
> required time-out periods in the specs, these should be left
> entirely up to the implementation as if done correctly these will
> likely be dynamically variable depending on the resource
> utilisation on the server, i.e. if a servers resources are becoming
> constrained it would decrease the time-out period depending on just
> how constrained it is and also if a server is not constrained and
> has plenty of resources going spare it would keep the connections
> open for longer and longer periods to avoid needing to set the
> connections up again.
The two things I would put in a specification somewhere is a mandate
that servers don't reconnect purely because of a disconnection, and
that "idle" is not related to link activity, but information flow.
(As in, "maintaining" a connection via XEP-0198 or XEP-0199 pings is
not possible). Of course, when really constrained, servers might want
to cut themselves off from the outside world entirely - intra-domain
communications will generally be more important.
The other half - figuring out if you missed something while
disconnected - is already specified in XEP-0198.
Dave Cridland - mailto:dave at cridland.net - xmpp:dwd at jabber.org
Infotrope Polymer - ACAP, IMAP, ESMTP, and Lemonade
More information about the Standards