Re: write-up of interaction patterns

Hi Sanjiva, all:

In general, I think we're all heading pretty much in the same direction here,
it's just a matter of tweaks and preferences.

As I see it you've got two things which can be (should be) named in an MEP
description.  First, the nodes themselves, and second, the messages which
travel between them.  For instance, in a request-response MEP, you might have:

    Node "A" sends message "Request" to node "B"
    Node "B" then sends message "Response" to node "A"

When you lay this into WSDL, I think you need both "what node am I" and "what
messages in the WSDL correspond to what messages in the MEP description.  An
example of this would be as follows (using my preferred syntax, natch :)):

<operation mep="request-response-uri" role="node-B-uri">
  <input message="someMessage" role="Request"/>
  <output message="otherMessage" role="Response"/>
</operation>

If you wanted to write the same WSDL from the perspective of the sender (node
"A") you could just flip it:

<operation mep="request-response-uri" role="node-A-uri">
  <output message="someMessage" role="Request"/>
  <input message="otherMessage" role="Response"/>
</operation>

Here's what I like about this:

1) it adds to the current syntax, rather than introducing yet another element
to the language.

2) it maintains the "input" and "output" distinction, which might (or might
not, I'm not 100% sure) be useful for certain tools, regardless of their
understanding of the MEP involved.

I like #1, but can certainly live without it as long as the same information is
conveyed in the "<interaction>" model - so you'd need to add the "role"
attribute to the <interaction> element to indicate which node in the MEP spec
is being represented by the WSDL.  That said, I still have a preference for
keeping <operation> and extending it.

As for #2, I think it could be a benefit, but we'd need (as with much of this
stuff) use cases to make sure.

Sorry this is a bit brief, but I hope it elucidates a bit more of what I was
proposing last week.

--Glen


----- Original Message -----
From: Sanjiva Weerawarana
To: www-ws-desc@w3.org
Sent: Monday, January 13, 2003 4:35 PM
Subject: write-up of interaction patterns


At the last call I was asked to write up the interaction pattern stuff
to try to get convergence. Document attached.

Glen, I don't think I did justice to your preferred approach. Please
edit as appropriate ..

Bye,

Sanjiva.

Received on Friday, 17 January 2003 11:40:26 UTC