RE: Let's fix THE PROBLEM

Shane Dempsey wrote:
> In my opinion it definitely makes sense to split 
> presence activities into Information (pdif?) and 
> Transport (PubSub events) concerns. 
	Layering of protocols and careful architectural decomposition
were once common practices in protocol design, however, these
disciplines seem to have been lost or forgotten in many of the
application areas in recent years. Where once people would seek to reuse
the implementation and design efforts that had been devoted to other
projects, the tendency today seems to be one of "reducing dependencies"
on other efforts. The result is that we get protocols like Jabber/XMPP
and SIP/SIMPLE that effectively rebuild an entire protocol stack and
provide new implementations of a large number of services that could be
provided by related protocols.
	Presence should definitly be split as you suggest.

> Arguably my toaster has as much right to produce 
> presence information as I do
	But, you will want more from your toaster than just "presence"
information! It is nice to know that the thing exists, however, you
probably would like to know what it is doing, what its current
temperature is, when it starts to process a new piece of toast, and
you'd like to be alerted whenever an internal sensor detects a
significant change in state. You could pump all this over some
"presence" system, but the reality is that little of this data is really
"presence" related. What you should have is a PubSub transport layer
that allows the toaster to send out messages that are of type "Presence"
or of type "Toaster Status". These are two types of message that use the
same PubSub transport layer (i.e. Subscribe, Publish, Notify, Watch,
etc...)
	Many of the "presence" applications that I hear about (other
than IM buddy lists) appear to involve exploiting the Presence protocol
to send information that isn't really "presence" related but would be
better handled by a PubSub protocol. What is happening is that people
are recognizing that the presence protocols are likely to be widely
disseminated and they are trying to figure out how to exploit these
implementations as transports for other applications. i.e. Presence,
which should be something that lives at the presentation layer, is being
treated as a transport layer protocol. This is wrong.

		bob wyman




  [reminder: [email protected] for non-technical discussions, please]