Re: Follow up on output ops + MEPs

Hey, all,

On 20 Dec 2002 07:45:35 -0500
Sanjiva Weerawarana <sanjiva@watson.ibm.com> wrote:

> On Thu, 2002-12-19 at 16:13, Prasad Yendluri wrote:
> > It seems you are proposing that we retain the <operation> construct of
> > portType to define the Web service interface (in WSDL 1.1 way) esp.
> > for input-output and input-only type operations and allow use of the
> > new <interaction> construct for (these plus) other patterns? Kind of
> 
> Yes pretty much - that the <interaction> construct is a general
> framework and that we have chosen to provide specific syntax for
> certain common interaction patterns. I'd keep the current operation
> syntax for input/output and input-only patterns only. If we do
> some pub-sub thing then I'd introduce new syntax for that too. Same
> for other forms of outbound ops we decide to define.

Speaking as an implementor, this looks horribly complex to me.  What do we gain by using these different element names over using a 'name' attribute?  I don't get it.  I don't understand why we want to distinguish between classes of exchange patterns, or what the criteria are for determining membership in a class.  I don't understand what the perceived difference between these classes is, so I can't see why it could be significant enough to require a new complexType.  I don't understand the criteria, so I don't understand under what circumstances, as a potential builder/publisher of extensions, I would use existing syntax versus creating my own.

I guess I just don't see the motivation for doing this.  Could someone provide motivation and classification criteria?

Amy!
-- 
Amelia A. Lewis
Architect, TIBCO/Extensibility, Inc.
alewis@tibco.com

Received on Friday, 20 December 2002 11:42:38 UTC