RE: [wsbpel] Issue - 166 - Proposal To Vote

[<andrew.francis@mail.mcgill.ca>]

Hi Bernard:

> an assign-specifying appendix sounds usefull to me.

Major Point:

Thanks. I believe a combination of a formalism (as
simple as pre-conditions, post-conditions) and
and getting the original author of section 14.3 to
state their intent, would have put section 166 to
rest a long time ago.

That said, too much of BPEL is English narrative
that is easy to misinterpret. Features like better
illustrations and pseudo-code would better explain
things. I also suspect it would shrink the size of
the specification.

> are some points, which are not really addressed. For
> example is the assign schema aware? If I have a "a (b,c)"
> complex type and assign
>
> <from>C</from><to query="/a/c" />
> <from>B</from><to query="/a/b" />

Minor Point One:

I would suspect that assign is schema aware because
variables and messages have type (and type is a XSD
schema/WSDL definition).

Minor Point Two. According to section 9.3, only partnerlinks,
variables, and variable properties can the target of
a <to>. Or in other words, a lValue. I suspect a parser
would (or ought) to identify the query as an expression.

I feel section 9.3 is somewhat arbitary. My feelings are
any mutable infoset (variables, messages) should be a
lValue candidate. Likewise XPath should be used to
navigate any BPEL object. This would allow messages
and variables to be treated more uniformily. Also
perhaps getVariableProperty() and getVariableData could
be depreciated.

I suspect something like XML Algebra has a better way
for expressing things like assignment and equality.
BPEL talks about equality but does not give a good
definition, or even a reference to its equality model.

Cheers,
Andrew