[standards-jig] Namespaces

David Waite mass at akuma.org
Thu Aug 8 00:02:07 UTC 2002


Robert Norris wrote:

>>The other thing which would be rather nice is if server connections and 
>>component connections sent messages and the like in the 'jabber:client' 
>>namespace rather than in the 'jabber:server' or 
>>'jabber:component:accept' namespaces. A message in the jabber:client 
>>namespace 'becomes' a message in the jabber:server namespace during 
>>transport, which doesn't matter now (since the namespace context of 
>>traffic from a client is not understood by the server), but will matter 
>>soon (once the server understands namespaces, it will have to rewrite 
>>every packet multiple times.)
>>    
>>
>
>Oh wow. I completely missed this one, and of course its correct. Hmm.
>
>[thinking out loud]
>
>How hard would it be, I wonder, to start gradually moving top-level
>packets into their own namespace, seperate from the stream "type"? The
>XMPP drafts already do this to some extent - the schemas for <message>,
><presence> and <iq> say that those elements are the
>'http://www.jabber.org/protocol' namespace.
>  
>
*gah* where did _that_ come from? They most certainly are _not_ in that 
namespace. Not saying I wouldn't mind if they were...

It is most probably an issue with the schema author.

>Then perhaps, if necessary, add an additional attribute or namespace to
>the stream that specifies the stream "type" - roughly analogous to the
>use of jabber:client/jabber:server/jabber:component:accept.
>
That works. As long as we are changing things, I would not be opposed at 
all to just using SASL as a single authentication mechanism and having 
the authentication indicate the authorization of the connecting entity 
to do special things, such as database and logging requests.

-David Waite




More information about the Standards mailing list