[Standards-JIG] Gateway Interaction (JEP-100) Comments

Thomas Muldowney temas at box5.net
Wed Mar 10 19:22:56 UTC 2004


Life is starting to stabilize and I've finally found time to really 
read this JEP.  So here are my first pass comments that I'd like to 
discuss, debate, whatever.

- Section 4.1.1 Step 6,  This should not specify that the gateway logs 
in, rather that the gateway checks that the credentials are valid.  If 
the gateway MUST login to do this, then Step 7 makes sense.
- Section 4.1.1 Step 7, This should specify that _all_ translatable 
events should be queued not just message events.
- Section 4.1.1 Step 8, What happens if the user denies the 
subscription here?  I don't think I ever handled this case, and it's 
very odd since you potentially have a bunch of events from Step 7.  It 
wouldn't make sense to just drop those once the user signs off would 
it?
- Section 4.2.1 Step 3, This might have similar problems as Step 8 
above.
- Section 4.4.1 Step 1, I think common practice for a long time was 
silently dropping probes unless they could immediately  return 
meaningful data, which is hard to do when you are offline.
- Section 4.4.1 Step 4, I feel the "Note" puts undo restrictions on the 
gateways.  What happens when the developers find they can better map 
users actions using resources?  Maybe they are in multiple groupchats 
that are only identified by userid, then resources could be used to 
fully identify what's happening.
- Section 4.5, This completely misses gatewayed users on the legacy 
system.  They need to have offline pres sent for them as well.
- Section 4.6.1 Step 1, This should probably be the gateways attempts a 
legacy subscription, and not just blindly move forward.  Some systems 
can tell if they are denied or not.
- Section 5.3.1 Step 1, Again, this "Note" does not sit well.
- Section 6, The loss of iq:gateway is odd to me.  It was simple and 
seems to do a job well, why not use it?  Point 0 against iq:gateway 
seems very short sighted, what happens if the other networks change 
before they make XMPP gateways?  What happens if other odd networks are 
gatewayed using this?  Point 1 again iq:gateway is largely valid, but 
could probably be solved in a nice JEP.
- Section 8 Point 0, I think I'm missing something, why is this 
necessary?
- Section 8 Point 1, Generally I think this fails in practice.  It's 
very very hard to distinguish what a JID is just based on the JID.  
Dunno what to say really.
- Section 9 Point 0, "sent in the clear" is largely false now, and 
almost a scare tactic, that should be removed.  The whole point could 
just go away.

There we go, a few questions in there I hope people chime in on.  It's 
almost there, but needs to clear a few things up first.

--temas




More information about the Standards mailing list