Re: [xmpp] probe from/to (was: Re: 3921bis: probe + unavailable)

[Waqas Hussain <waqas20 at gmail.com>]

On Fri, Jan 29, 2010 at 8:03 PM, Peter Saint-Andre <stpeter at stpeter.im> wrote:

On 1/29/10 3:37 AM, Waqas Hussain wrote:

> There's another thing I was wondering about:
>
> For presence probes, "The value of the 'from' address MUST be the full
> JID <user at domain/resource> of the user and the value of the 'to' address
> MUST be the bare JID <contact at domain> of the contact to which the user
> is subscribed."
>
> Why MUST those JIDs be that way? I can think of valid cases where a bare
> JID may wish to probe the presence of a full JID.

The original reasoning was that a user comes online with a particular
resource and my server sends a probe to the user's contacts (bare JID)
on behalf of that resource (full JID). I don't know what your use cases
are, so I'm not sure why we would change the logic we've had since 1999.

I was concerned about presentities which are not users. If component.example.com wishes to probe user at example.org, MUST it send the probe from component.example.com/someresourcewhichisntactuallyusedforanythingelse?

And there are valid cases where sending to a full JID would be the optimal solution. A MUC can use it to detect ghosts. And no, it doesn't have to poll every minute. If it has information about S2S for example, a smart implementation could just poll when it has detected that the remote server reconnected, restarted, etc, which is better than no ghost detection. The usual way to test for a ghost is to detect remote-host-not-found errors. And if you missed that (e.g., you didn't send any stanzas while the remote host rebooted), then is there any other way to detect ghosts which works with existing clients?

My only objection is that MUST is too forceful.

--

Waqas Hussain