Re: [wsbpel] Draft: Issue - 11 - Discussion & Proposals

[Ron Ten-Hove <Ronald.Ten-Hove@Sun.COM>]

Rania Khalaf wrote:

When we did 157, xslt was proposed and a lot of people - or maybe a few very vocal people (i can't recall who right now) - were very much against xslt since it 'would introduce lots of unknowns and we don't have time to debug this now' or something like that.

so my point is, that if we go the xslt way here, that affects 157 as well and the arguments against xslt should be discussed.

The 157 proposals would have introduced a fairly tight coupling between the BPEL copy mechanisms and an XSLT-based transformation. This indeed would have been complex, since the intermixing of the two vocabularies would have had a lot of interesting (read complex) nuances.

This proposal keeps XSLT at arms-length. The simplest variant, as an XPath extension function, doesn't impact the BPEL data handling model at all.

i personally like chris's proposal and especially if uses the infoset language as yaron had suggested.

The motivation here is to give us a portable way of putting non-trivial XML manipulation into BPEL processes.

By using the XPath extension function variant, we don't need to define the mechanism in terms of info-set. If we go for the extensibleAssign version, that would require proper info-set definitions.

-Ron