[Standards-JIG] Re: AMP & deliver condition

Justin Karneges justin-keyword-jabber.093179 at affinix.com
Mon Jul 5 19:15:56 UTC 2004


On Monday 28 June 2004 8:39 am, Peter Saint-Andre wrote:
> On Fri, 04 Jun 2004 18:04:25 -0700, Justin Karneges wrote:
> > In the past, I discussed the Jabber "black hole" problem:
> >
> >   http://www.jabber.org/pipermail/standards-jig/2003-December/004589.html
> >
> > It appears as though AMP might be able to solve some of the issues.
> >
> > 1) Can the deliver condition be used so that a client knows his sent
> > message at least reached his own server (first hop) ?
> >
> > 2) Can the deliver condition be used by a server to a client, so that the
> > last server in the chain knows if his client received it?
> >
> > 3) Can the deliver condition be used for full end-to-end acking, as a
> > replacement for JEP-0022?
>
> Exactly what problem are we trying to solve here? Assurance that a message
> was (1) received by a client or application, (2) displayed to a user,
> or (3) actually read by a user or processed by an application? To me, it
> seems that *perhaps* we're trying to use (1) or (2) as a proxy for (3).
> But my client could affirm that (1) and (2) happened, yet I still never
> read the message because the BSOD intervened or somesuch.

None of the above.  The problem is simply to determine if a stanza has made it 
across a single hop, such as client-to-server.

The "black hole" effect, as described in the quoted URL, can arise when the 
user tries to send a message to his server after it has gone down.  During 
this period of downtime (somtimes up to 20 minutes, depending on the 
underlying TCP/IP configuration), all stanzas sent to the server will go into 
the bit bucket, and there is no easy way for the client to know this.  JEP-22 
"delivery" events are not a complete solution, since they rely on the remote 
entity being around to reply, which is too far-reaching for what we want to 
accomplish.

-Justin




More information about the Standards mailing list