+        if xpath.matches("/presence/decloak", stanza):

Check the xmlns, please.

+        # FIXME: if a client sends a directed unavailable after they make a
+        # call, currently we just log out. What we should probably do instead
+        # is only listen to presence update/subscribe on the gateway JID and
+        # ignore presence to virtual users for logging in/out.

Er, yes, that. I predict that to handle this sort of thing correctly, we'll end up with two completely separate sets of callbacks: one for user@gateway, and one for the gateway itself.

+        # Also, this identity is shared by our virtual users. Hope this is ok.

I think the gateway and the individual users should have distinct caps and perhaps also identity. After all, you can't Jingle-call <gw.example.com>, only <bob%example.net@gw.example.com>, so gw.example.com shouldn't have Jingle caps.

+        # NOTE: We assume here that they are querying us for our caps, and
+        # therefore don't bother checking that the caps are the same.

Um...

+        c['node'] = 'http://telepathy.freedesktop.org/wiki/fargo'

/Fargo has the benefit of actually existing :-P

> +import ns

Absolute imports, please: import telepathy_fargo.ns as ns. Rationale: http://www.python.org/dev/peps/pep-0328/

(Perhaps we should switch on "from __future__ import absolute_import" in all modules, since AIUI we require Python 2.5 anyway?)

Do you want to merge up your branch and I'll do these fixes myself?

Sorry, those comments were regarding <http://git.collabora.co.uk/?p=user/alsuren/telepathy-fargo.git;a=shortlog;h=refs/heads/caps>.

(In reply to comment #1)
> Check the xmlns, please.

Fixed in my caps branch.

> 
> +        # FIXME: if a client sends a directed unavailable after they make a
> +        # call, currently we just log out. What we should probably do instead
> +        # is only listen to presence update/subscribe on the gateway JID and
> +        # ignore presence to virtual users for logging in/out.
> 
> Er, yes, that. I predict that to handle this sort of thing correctly, we'll end
> up with two completely separate sets of callbacks: one for user@gateway, and
> one for the gateway itself.

Done for IQs in my caps branch; still to be fixed for presences.

> +        # Also, this identity is shared by our virtual users. Hope this is ok.
>
> I think the gateway and the individual users should have distinct caps and
> perhaps also identity. After all, you can't Jingle-call <gw.example.com>, only
> <bob%example.net@gw.example.com>, so gw.example.com shouldn't have Jingle caps.

Still to be fixed.

> +        # NOTE: We assume here that they are querying us for our caps, and
> +        # therefore don't bother checking that the caps are the same.
> 
> Um...

To be fixed.

> +        c['node'] = 'http://telepathy.freedesktop.org/wiki/fargo'
> 
> /Fargo has the benefit of actually existing :-P

Fixed in my caps branch.

> > +import ns
> 
> Absolute imports, please

Fixed in my caps branch.

Some other problems related to presence/disco/caps handling:

* TelepathyService.got_iq assumes that @to will either be the gateway or one of our virtual users, not some other address (just FIXME'd for now)

* needs regression tests!

* presence probes aren't implemented (Bug #26141)

This branch still doesn't address my complaints, but it fixes some other nits regarding caps, and adds the beginning of regression test coverage for presence and probes.

http://git.collabora.co.uk/?p=user/smcv/telepathy-fargo.git;a=shortlog;h=refs/heads/caps

(In reply to comment #3)
> (In reply to comment #1)
> > Er, yes, that. I predict that to handle this sort of thing correctly, we'll end
> > up with two completely separate sets of callbacks: one for user@gateway, and
> > one for the gateway itself.

Fixed in master.

> > I think the gateway and the individual users should have distinct caps and
> > perhaps also identity. After all, you can't Jingle-call <gw.example.com>, only
> > <bob%example.net@gw.example.com>, so gw.example.com shouldn't have Jingle caps.

Fixed in master.

> > +        # NOTE: We assume here that they are querying us for our caps, and
> > +        # therefore don't bother checking that the caps are the same.
> > 
> > Um...

Bug #26216.

> > +        c['node'] = 'http://telepathy.freedesktop.org/wiki/fargo'
> > 
> > /Fargo has the benefit of actually existing :-P

Fixed in master.

> > > +import ns
> > 
> > Absolute imports, please

Fixed in master.

> * TelepathyService.got_iq assumes that @to will either be the gateway or one of
> our virtual users, not some other address (just FIXME'd for now)

Bug #26204 (which is just for robustness - the host ejabberd shouldn't forward stanzas that would fail that check anyway)

> * needs regression tests!

Fixed in master.

> * presence probes aren't implemented (Bug #26141)

Fixed in master.

I'm going to consider this done, modulo the parts that have been spun out as separate bugs.