[Standards] publish+configure again

[Pedro Melo]

On Apr 2, 2007, at 8:36 PM, Tomasz Sterna wrote:

> Dnia 02-04-2007, pon o godzinie 15:24 +0100, Pedro Melo napisał(a):
>> Of course. The thing is, PEP or PubSub already have a lot of uses
>> cases covered. That's why it seems the best approach to use pep to do
>> private storage.
>
> I always advocate "use proper tools for proper things" approach.
> That's why I do not feel well looking at the proposal of making a
> distributed storage system of PEP/PubSub.
>
> Let's store things in a storage system and use eventing system to
> propagate events.

Sorry, but the moment that PEP keeps the last item (at least) to send  
to clients that connect later on, PEP is a storage system, and you  
need to think about the security considerations of that.


> I do not liked the persistent nodes idea from the very beginning.  
> PubSub
> was distributing events in <message/>s storable in offline storage for
> later retrieval. There was no need for persistence in PubSub.

Forcing just-connected buddies to wait for the next event to get a  
clear picture of my status will only force the existence of new  
protocols, because a lot of people will want to know immediately my  
user tune, or my user geo-location or my mood.

I prefer to have a single protocol, PEP, that solves both problems.

Another devious side effect of not having storage in PEP would be  
(I'm speculating here) that client would publish more often to lower  
the delay between updates, and that seems worse than storage.

> And the possibilities of abuse... infinite.
> There are applications that allows storage of arbitrary amount of data
> in DNS caches that are very volatile. With persistent PEP nodes it's
> even much simpler.

I don't see how... DNS caches trust responses from several sources.  
PEP only trusts publish items from a single local user. I don't see  
how you can make this claim.

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