[Standards-JIG] Closing idle streams

Dave Cridland dave at cridland.net
Thu Jun 1 23:05:04 UTC 2006


On Thu Jun  1 23:50:31 2006, Peter Saint-Andre wrote:
> The question is: can a server end its session with another server 
> on a
> whim, or does it always need a reason? Theoretically I tend to feel 
> that
> it would always be friendly to specify a reason, but I'm not 
> convinced
> that in practice that's really necessary.

I agree with your theory, and in practise I don't think it's a 
significant overhead in most cases. I'd call it a RFC2119-SHOULD. I'm 
somewhat neutral in whether a JEP is needed here or not, but I do 
think it belongs in 3920bis.


Now... I said "in most cases", and SHOULD rather than MUST. The cases 
where I think it's be good to close the stream without a reason are:

1) Where there is a significant cost-per-octet. I can't really see 
this being the case for s2s connections.

2) Where there is very low available bandwidth. This could be the 
case for s2s connections, when they're simply running on a choked 
link or something, or when they're running in largely isolated 
networks.

3) Where the link forces a very small packet size. I can't see this 
being a real issue.

Otherwise, the error and stream closure can both fit in a single 
packet, and the overhead isn't big in percentage terms.

Dave
-- 
Dave Cridland - mailto:dave at cridland.net - xmpp:dwd at jabber.org
  - acap://acap.dave.cridland.net/byowner/user/dwd/bookmarks/
  - http://dave.cridland.net/
Infotrope Polymer - ACAP, IMAP, ESMTP, and Lemonade



More information about the Standards mailing list