[Standards] Jingle and invisibility

Lauri Kaila lauri.kaila at gmail.com
Tue Jun 5 00:52:23 CDT 2007


2007/6/5, Peter Saint-Andre <stpeter at jabber.org>:
> Lauri Kaila wrote:
> >
> > XEP-0166 says that the session should be ended when presence of type
> > "unavailable" is received from the other party. What if the reason was
> > that the other party became invisibe? Session could stay alive, unless
> > IQ was also blocked. What if an initiation is received from an
> > invisible resource?
>
> Good point. In general, it is probably best not to tie the Jingle
> session to presence so closely, so I may remove that note anyway. The
> only event that must be considered session-ending is receipt of a
> "session-terminate" message. If the other person goes offline by sending
> presence of type unavailable, you'll find out soon enough if the session
> is over (bounced packets etc.). You may consider the session ended at
> that point, but I think that's optional.

I first thought that presence unavailable + no Ping (0199) response
would be a good indication that the session is gone. Then client must
implement some timer for waiting for the Pong. Also it leaves an
unlikely case for false positive. For example, client crashed and then
quickly got up with the same resource name. So it isn't worth for the
effort. As you said, the same information is possible to get from the
media channel.

However, I think clients should be allowed to use all kinds of hints
to determine automatically that the session is gone, but then they
should send session-terminate just in case.

BTW, speaking of XEP-0199: How long one should wait for a Pong?

-lauri


More information about the Standards mailing list