[Standards] pubsub/pep auto-creation

Pedro Melo melo at simplicidade.org
Thu Mar 22 12:02:02 UTC 2007


Hi,

On Mar 22, 2007, at 7:49 AM, Remko Tronçon wrote:

>> so in practice it seems to me that clients are just always going  
>> to do
>> publish+configure.
>
> Indeed, that's the idea.
>
>> And that seems a bit excessive to me.
>
> Why? Rachel pointed out that, if you don't do that, you have to check
> the configuration *every* time you publish an event, and possibly
> modify the node. It gets even worse: between the check and the
> publish, some other resource might have made your node public for a
> public item it is publishing, and you're publishing something private
> to a public node.
>
> I'm still convinced that specifying the 'how' (public/private)
> together with the 'what' (event) is easier to grasp, easier to
> implement, and avoids a lot of trouble. I'm also throwing in my
> standard excuse of 'stream compression will take care of the extra
> traffic' (although actually, it takes a lot more traffic to publish
> events 100% correctly if you have to check configuration every time).

Also, a lot of non-XMPP publish/subscribe systems work that way: the  
sender specifies the behavior that the item published should have.

In XMPP this means sending the configuration every-time because we  
don't have the notion of a publisher-session.


[.. two relevant examples delete, specially the backstreet boys  
thing ..]

> But all in all, the convincing argument should be the fact that you
> can never assume on a publish that you are publishing an item the way
> you want it, unless you check configuration. And even checking doesn't
> guarantee you that your item will be published privately if some other
> resource is posting public events to it.

Exactly. And this is pretty important. +1 for publish+configure.

The only alternative would be to create the notion of "publisher  
session", where you would configure when you login, and the PEP  
service would remember the configuration you set for that node, and  
whenever the same full jid publishes to that topic, the PEP service  
would assume the configuration sent at the beginning. Although this  
might work, I think it places an unnecessary burden on the PEP  
component.

Best regards,
-- 
Pedro Melo
Blog: http://www.simplicidade.org/notes/
Jabber ID: melo at simplicidade.org
Use Jabber!





More information about the Standards mailing list