RE: [wsbpel] Issue 11 - Call for Discussion

["Kristofer Agren" <kagren@pakalert.com>]

How about coming up with some construct that allows for variables to be
manipulated within a certain context where the variable contents are not
validated until the end of that context?

Perhaps adding an attribute to the "scope" element that specifies whether or
not variable contents are validated within the scope? For example:

<scope name="outerNormalValidatingScope">
	<variable name="test" type="a:type"/>
	<sequence>
		<scope name="modifyTest" validateVariableContent="no">
			<while ..>
				Build variable contents incrementally here
			</while>
		</scope>
	</sequence>
</scope>

Kristofer

-----Original Message-----
From: Ugo Corda [mailto:UCorda@SeeBeyond.com] 
Sent: Tuesday, March 02, 2004 9:39 PM
To: edwink@collaxa.com; ygoland@bea.com; Alex Yiu
Cc: Ron Ten-Hove; Wsbpel@Lists. Oasis-Open. Org (E-mail)
Subject: RE: [wsbpel] Issue 11 - Call for Discussion

> 3. As pointed out, this is orthogonal to type validation 
> because some use
> case: {while+invoke+receive+assign} require the incremental 
> build out of documents. It seems that a separate issue should 
> be opened to track validation (which could be addressed by 
> doing validation only on the edges). More generically, any 
> logic that combines sync or async interaction and build out 
> of a document will not be able to be compressed into a single 
> activity, companion language or not.
> 

That's the way I see it too. The validation problem in the
{while+invoke+receive+assign} loop does not come from the assign itself
but from the incremental building of the XML typed variable through
successive loop cycles. No matter what we substitute the assign with,
the problem remains. The bottom line is that it is only at the
completion of the loop that it makes sense to validate the variable
contents.

The only alternative approach seems to be encapsulating the entire loop
into a separate Web service (with the assumption that whatever we use to
implement that Web service does not insist in doing validation before
the end of the loop). But that means that BPEL loses control of the
detailed operations that occur inside the loop itself, which in many
cases might not be what is desired.

Ugo


To unsubscribe from this mailing list (and be removed from the roster of the
OASIS TC), go to
http://www.oasis-open.org/apps/org/workgroup/wsbpel/members/leave_workgroup.
php.