W3C home > Mailing lists > Public > public-ws-chor@w3.org > September 2004

For discussion on tonight's agenda (WSDL and Interacts)

From: Steve Ross-Talbot <steve@enigmatec.net>
Date: Mon, 20 Sep 2004 15:19:25 +0100
To: WS-Choreography List <public-ws-chor@w3.org>
Message-Id: <1387D639-0B10-11D9-A1F1-000393D13C9A@enigmatec.net>
Main agenda item will be to get a common understanding of WSDL and 
interaction in WS-CDL.
I shall send a formal agenda later today. I do not intend to ask for 
any decisions and infact will rail
against doing so in the absence of so many people from the call.

To fuel the discussions for tomorrow I include the following:

<start>
In the last document I have for the spec (dated 16th August) it says:

<start>
2.5.2 Interaction
An Interaction is the basic building block of a Choreography, which 
results in the exchange of information between parties and possibly the 
synchronization of their states and the values of the exchanged 
information.
An Interaction forms the base atom of the recursive Choreography 
composition, where multiple Interactions are combined to form a 
Choreography, which can then be used in different business contexts.
An Interaction is initiated when a party playing the requesting Role 
sends a request message, through a common Channel, to a party playing 
the accepting Role. The Interaction is continued when the accepting 
party, sends zero or one response message back to the requesting party. 
This means an Interaction can be one of two types:
A <emph>One-Way Interaction </emph>that involves the sending of a 
single message
A <emph>Request-Response Interaction </emph>when two messages are 
exchanged
</start>

First of all this talks about a One-Way *Interaction* and a 
Request-Response *Interaction*. It does not specify a WSDL MEP anywhere 
in the text. There is also nothing described about how faults are 
defined for interaction and where they might come from. Thus nothing at 
the WSDL fault level is described which ties CDL to WSDL of any 
flavour.

Given this is the case there is nothing to restrict the mapping of 
these *Interaction* patterns to the existing Interact language 
construct. A notification maps to a One-Way interaction (it is after 
all a single message). A solicit-response which is the reverse of 
request-response (it is after all two exchanged messages) to a 
Request-Response interaction which is just a pair of exchanged 
messages.

I would suggest that the other MEP's in WSDL2.0 are variants of these 
and can be dealt with in a similar way.

Thoughts?

I have a service A1 and B1. A1 sends a request to B1 and B1 responds. 
Imagine an interact
that describes this.

Now supposing A2 and B2 do the same thing but B2 does something else 
after it receives
the request which is hidden from A2 (not in the choreography). Then the 
request/response
are async as opposed to that between A1 and B1 which may be sync.

In my view the choreography, the interaction that describes this, does 
not change. All that
changes is the bindings to the service definitions since one is sync 
and the other async.

Does this make sense too?

I enclose a note from Gary with further information/questions/examples:



</start>


Received on Monday, 20 September 2004 14:19:41 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 18 December 2010 01:01:05 GMT