[standards-jig] Pub/Sub Explanation

Piers Harding piers at ompa.net
Wed Apr 17 07:56:38 UTC 2002


To me, there is a recurring problem in how people view the question of how to deal with the data that the end user/consumer has subscribed to.

The thing that troubles me, is that the simple fact of what that data actually means is application specific ie. a "contract" between the publisher and the consumer.  This is not being recognised for what it is.  As far as the pubsub protocol, and its implementations are concerned it shouldnt give a "monkeys" what that data is.

What counts is how the consumer interprets that data ie what it "means" when:
the data is there or not ( in the payload )
when the data has changed or not
what the payload contains
or even if the consumer cares this time round when notified

... This is all up to the publisher and the consumer to figure out, directly.

Once we get past this point, we can then use it to compartmentalise a number of the major issues, such as data storage ( again something for the publisher and consumer to negotiate eg. whether it is in the payload, or the payload is a pointer to the location of the data etc. ), and how to identify, and divulge meaning from a topic identifier - ie. topic management, and categorisation.

Tell me I'm wrong :-).

Cheers.


On Tue, Apr 16, 2002 at 02:27:31PM -0500, Peter Saint-Andre wrote:
> Peter Millard and I were chatting over lunch, and he mentioned that the
> initial motivation for developing pub/sub systems in Jabber were things
> like this:
> 
> 1. I publish my vCard
> 2. I publish my public key
> 3. I publish my avatar
> 
> These are things that you might want to subscribe to (i.e., know when they
> change, which change is an event), and that we'd probably want to be
> persistent.
> 
> Peter
> 
> --
> Peter Saint-Andre
> email+jabber: stpeter at jabber.org
> weblog: http://www.saint-andre.com/blog/
> 
> On Tue, 16 Apr 2002, DJ Adams wrote:
> 
> > On Tue, Apr 16, 2002 at 11:59:52AM -0500, Peter Saint-Andre wrote:
> > 
> > >[...]
> > > we have pure pull, pure push, or hybrid depends on what kind of data
> > > is being published. Think of a simple pub/sub system for weblogs. I
> > > could be notified when your blog changes and then I have to pull down
> > > the URL for your latest story. I could be notified when your blog changes
> > > but also receive an abstract of the story, thus helping me decide if I
> > >[...] 
> > 
> > > Well, those are just random thoughts, hopefully someone finds them of use.
> > 
> > Sure!
> > 
> > In fact, I've experimented with both scenarios (a payload-less) event
> > type scenario (that's what we're doing with Radio right now), and a
> > publish scenario with Peerkat, where the whole item is published.
> > 
> > I don't think either of these two scenarios are in question in either
> > case; I think it's more related to what mechanism is looking after any
> > data that might be polled (in the Radio case, for example, it's a
> > nation of an RCS (Radio Community Server), and the Radio desktop itself. 
> > 
> > Cheers
> > dj
> > _______________________________________________
> > Standards-JIG mailing list
> > Standards-JIG at jabber.org
> > http://mailman.jabber.org/listinfo/standards-jig
> > 
> 
> _______________________________________________
> Standards-JIG mailing list
> Standards-JIG at jabber.org
> http://mailman.jabber.org/listinfo/standards-jig



More information about the Standards mailing list