[Standards] Namespace Priority

Ian Paterson ian.paterson at clientside.co.uk
Tue Mar 13 00:27:03 UTC 2007


Peter Saint-Andre wrote:
> Ian Paterson wrote:
>> There seems to be a real need. We've already got servers like Tigase 
>> implementing proprietary functionality. Shouldn't we at least 
>> standardise something? Why shouldn't we define a protocol (perhaps 
>> something like this: 
>> http://mail.jabber.org/pipermail/standards/2007-February/014000.html) 
>> to enable those applications that want to support specialized content 
>> routing?
>
> Can't the server already figure that out based on RAP? It could use 
> the RAP info for routing stanzas sent to the bare JID even in cases 
> where the contact does not receive the user's presence. But it's not 
> clear how that applies to IQs, if at all.

It doesn't apply to IQs. But, as you mention below, protocols like 
Jingle can still use IQs because a client can first perform a Session 
Negotiation (XEP-0155 or XEP-0116) in order to discover the resource of 
the other entity before sending an IQ.

A server could use the RAP elements it receives in presence stanzas from 
its users to understand which namespaces each connected resource 
supports. However RAP elements are currently not designed for that. 
Furthermore, it doesn't help anyway, since servers would still have no 
way of understanding to which resource a Session Negotiation request 
should be delivered (since the sender currently has no protocol to tell 
the server that the request should be sent to a client that supports, 
for example, Jingle).

RAP only solves the issue for people who I allow to subscribe to my 
presence. So IMHO the solution should not involve presence stanzas at 
all. The alternative proposal outlined in the URL above does give 
servers a bit more work when delivering <message/> stanzas to bare JIDs, 
but it has several advantages:

1. it works for non-subscribers
2. it keeps presence stanzas small
3. it is even more simple for clients

> Negotiate a chat session via XEP-0155 to request an audience, when it 
> starts exchange directed presence, then do your IQ stuff for Jingle. 
> Seems straightforward to me.

How can the initiator (who is not a subscriber) tell the other user's 
server that its XEP-0155 session request should be delivered to the 
resource that supports Jingle? Even after the initiator discovers that 
the resource the request was sent to does not support Jingle, it still 
has no protocol to tell the server to deliver any subsequent request to 
a different resource.

- Ian




More information about the Standards mailing list