Presence and IntComm WG

[Xiaotao Wu <>]

On Tue, 17 May 2005, Candace Holman wrote:

> Thanks Xiaotao.
>
> Thanks for the clarification.  For n PUAs in p2p an update creates n-1
> notifies but in pa an update creates 1 publish plus n-1 notifies.  It
> doesn't count as a reduction because the traffic is only reduced locally, 1
> publish vs n-1 notifies.  It reappears elsewhere in the network.
>
> So, how about the multi-PUA case - if I have x devices/PUAs/accounts for
> one user, and user presence changes, does that send one publish to the pa,
> but in p2p mode it sends x notifies to each watcher/peer?

Presence change on each PUA may cause a PUBLISH to the PA, and the PA may
send a NOTIFY out (the tuple for that PUA get changed). This depends on
the policy on the PA. If the PA considers the presence status at user
level (e.g., only onlie/offline status for each user,  not sending device
information), the PUA status change may not trigger a NOTIFY. If the PA
also handles device status change for notification, each PULISH will cause
a NOTIFY out, and the number of NOTIFY will be the same compared to the
centralized model. But the number of subscription definitely get reduced.

-Xiaotao

>
> Candace
>
> At 03:49 PM 5/17/2005, Xiaotao Wu wrote:
> >inline.
> >
> >On Tue, 17 May 2005, Candace Holman wrote:
> >
> > > In anticipation of this week's conference call, Ben and I have been
> > > discussing some of the advantages of using a presence agent over using 
> > > peer
> > > to peer communication.  I question some of the advantages that I came 
> > > up,
> > > so please comment on this list.  Also, please add to this list if you 
> > > know
> > > of additional advantages.
> > >
> > > 1. Reduced traffic, more scaleable as number of peers rises
> >
> >I would say 'reduced traffic as number of each user's end devices rises'.
> >If every user has only one PUA, there is no traffic reduction (in fact,
> >more traffic because the PUA needs to send PUBLISH to the PA).
> >In terms of scalability, why a centralized model is more scaleable? In
> >fact, if every user has only one PUA, as the number of peers rises, P2P is
> >more scalable.
> >
> > >
> > > 2. Reduced access control administration (traffic comes from single 
> > > source)
> > >
> > > 3. Synthesized multi-ua information (presence update merging when 
> > > multiple
> > > devices are used by a single presentity)
> > >
> > > 4. Identity assurance (not strongly authenticated, but some assurance of
> > > intradomain identity)
> > >
> > > 5. Centrally managed contact lists and blacklists (not yet in PIC-SER)
> > >
> > > 6. Partial presence publishing (not yet in PIC-SER, not sure this has 
> > > to be
> > > centralized either)
> >
> >Yes, this has to be centralized. Each PUA uses PUBLISH to send its status
> >to its PA and the PA do the composing.
> >
> >Thanks!
> >
> >-Xiaotao
> >
> > >
> > > Thanks,
> > > Candace
> > >
> > >
>
>