Re: Stateful services (was Web Service Description and stateful s ervices)

* Champion, Mike <Mike.Champion@SoftwareAG-USA.com> [2003-06-24 11:08-0600]
> Umm, I'm losing track of this thread, at least as it relates to the Web
> Services Architecture WG.

Me too.

> I'd appreciate a bit more focus on what the WSA document should say
> about statefulness, and relating the concepts/relationships being
> kicked around here to those in the WSA.  Can someone summarize what
> this discussion has taught us about these things?  

Here is my attempt at it. Please feel free to correct me.

It seems that there was a lot of terminology used in lots of different
ways in the thread. I have tried to stay consistent with what seemed
to be the most often used.

There are two types of "state":
(1) the state of the service / resource / application (targetResource).
(2) the state of an interaction, which could involve one or more
participants.

A good illustration of this is at [1]. [2] was arguing three types of
states, but I think that from the point of view of our concepts, [2]'s
#1 and #3 are (1) above; I see the agent's and the data's state as
targetResource, or internal to the agent in which case we don't care
about it.

(1) is modified and represented via the Web service interfaces,
modified by the agent.

Creating and maintaining (2) is achieved by certain mechanisms:
- an identifier.
- at the communication/transport layer. Note that HTTP and REST do not
  allow you to maintain state like this: the client-server
  interaction is stateless.

The rest of the discussion was about expressing (2) in choreography
languages, but I am not sure what conclusions (if any) were reached
here. Maybe somebody who participated to this thread will have a
better view of it.

This makes me think that we need:
- a subsection about state in the stakeholders' section.
- to clarify the agent / targetResource relationship.
- to work on what needs to be expressed and how to do so.

Additional personal note: one way to achieve (2) is with a feature
(sometimes in a SOAP envelope, sometime accomplished by the underlying
protocol binding); complementary ways exist, e.g. by having different
resources for different interactions.

Regards,

Hugo

  1. http://www.w3.org/mid/20030620102144.U30095@www.markbaker.ca
  2. http://www.w3.org/mid/BC28A9E979C56C44BCBC2DED313A447001D757E2@bond.ncl.ac.uk
-- 
Hugo Haas - W3C
mailto:hugo@w3.org - http://www.w3.org/People/Hugo/

Received on Wednesday, 25 June 2003 06:59:43 UTC