[standards-jig] UPDATED: IBB (JEP-0047)

Peter Saint-Andre stpeter at jabber.org
Wed Mar 26 16:36:01 UTC 2003


On Tue, Mar 25, 2003 at 09:38:06PM -0700, Peter G. Millard wrote:
> ----- Original Message ----- 
> From: "Peter Saint-Andre" <stpeter at jabber.org>
> > > According to temas, removing the 'acks' would speed up the data
> exchange.
> > > This is why a sequence number was added.  If we were acking everything,
> then
> > > we wouldn't need such a counter.
> > XMPP Core (section 7.5.1) *requires* the ack:
> >       An entity that receives an IQ request of type 'get' or 'set'
> >       MUST reply with an IQ response of type 'result' or 'error'
> >       (which response SHOULD preserve the 'id' attribute of the
> >       request).
> > Peter
> 
> According to the JEP, all of the data transfers are sent using the same ID,
> and are all IQ-sets. I would argue, that the receiving entity responds with
> an IQ-result when it receives the iq with the <close> element and after it's
> received all of the sequenced packets. This makes more sense to me since all
> of the individual iq-sets are just part of a larger "transaction".
> 
> I'd like to see that response added into the JEP and the semantics around
> whether or not that stanza is a result or an error. Given the nature of
> delivery in jabber, it's unlikely that the receiver would not recieve one of
> the stanzas with data, but I suppose it's possible. There should be some
> errors around this.

This sounds agreeable to me.

Actually, it's been pointed out to me that right now there is no
expected result/error "ack" sent from a client in response to a roster
push. So MUST in XMPP Core may be a bit strong....

Peter

-- 
Peter Saint-Andre
Jabber Software Foundation
http://www.jabber.org/people/stpeter.php



More information about the Standards mailing list