Minutes
Attendance - 7 out of 19 - quorate
Approval of minutes
Minutes from last meeting approved
Resolution: Minutes of 2008-09-11 located at http://lists.oasis-open.org/archives/sca-bpel/200809/msg00011.html approved
AI review
AI-21 - Martin to follow up with Mary. Martin communicated with Mary: No response from Mary.
AI-46 - Mike Rowley not on the call. Still open.
Issue 18
Martin reviews what he wrote.
We looked at all the conbinations of things that may or may not exist.
Case 1 - Simple BPEL process, what's the component type
Case 2 - Extended with SCA extenstions
Case 3 - You have a CT side file that modifies what the component type looks like
No term for introspected or effective or derived CT. Need to define term.
After deployement what does effective CT mean?
<Dieter Koenig>
just looked at sca-j - there are several places talking about an "implied component type" - might be something that should
be consistent across specs
This is not so much abt conformance as rules == BPEL must look like this to get a good CT file.
Danny:
We shd have some grammatical productions that say what the rules are
Perhaps in a separate section well marked
<anish>
here is an example from the XML spec:
<anish>
[Definition: A software module called an XML processor is used to read XML documents and provide access to their content and
structure.] [
<Mike Edwards>
I still think that it is perfectly possible to write test cases for the introspected component type
<Mike Edwards>
clearly those testcases use Assembly + BPEL, but I don't see a way of avoiding Assembly in any testcase
Anish:
Clearly separate definitions from conformance criteria
<Martin C>
mike, i agree, but dont you do this by looking at what the runtime did or dodnt do
<Mike Edwards>
you define a BPEL process and a composite using that process as an implementation and expect it to either run successfully
or not, depending on the testcase involved (+ve or -ve test)
<Mike Edwards>
the test target is clearly the runtime
Martin:
Take a stab at my action item and see what happens
<Mike Edwards>
the overall combination of Assembly + BPEL impl is what gets tested
<anish>
mike, the test target will always be the runtime. one can't never say that your introspected CT was non-conformant
Sanjay:
Let's look at yr proposal in this TC first before taking it to Assembly
<Mike Edwards>
Anish: the effects of the introspected CT are testable - and that is all that matters
<Mike Edwards>
(I am reminded of quantum theory in that statement)
<anish>
i think we are agreeing. the effects are testable. the artifact is not.
Danny:
Having the BPEL file as a conformance target does not tell us what to do if it not a valid SCA file
Martin:
We could say SCA file or SCA Bpel file
<Mike Edwards>
A simple statement in the BPEL spec along the lines: IF the BPEL process document is invalid (wrt the schemas) THEN the runtime
must reject it
<anish>
or MUST not deploy the component/composite that uses an implementation that points to the bpel process
<Sanjay>
Mike, should we test the SCA specs by injecting them into the Large Hadron Collider?
It may find the specs impenetrable!
<Mike Edwards>
Hey - the LHC as an SCA test runtime - sounds great
<anish>
you have to collide two composites at a speed close to that of light to split it into components and examine if the component
points to the bpel process
Martin:
I have AI to rewrite section 2.1
Danny van der Rijn: LHC is a C&I
<anish>
may be the introspected CT is the 'god' particle
<Mike Edwards>
"total destruction" testing
Sanjay:
Martin will write prose for issue 18
<anish>
hot off the press: introspected CT behind dark matter and dark energy
Mike E:
Could use the approach we used for Java. 2 sections: one on unannotated POJO and the other on annotated
Anish:
I have question abt relation between CT side file and introspected CT
<Mike Edwards>
No - complete override should not be allowed
<Mike Edwards>
the BPEL Process is what it is
<Mike Edwards>
effectively, all you can do is restrict
<Mike Edwards>
you can't add a service
<Mike Edwards>
you can't remove an Intent
<Mike Edwards>
What do you mean by "tweak"
Mike E:
I'm nervous abt changing a service into a reference
<Mike Edwards>
I would not require the side file to reproduce the introspected CT
Danny:
Side file shd have all the power of BPEL annotation
<Mike Edwards>
but to remove just one service would require the CT to list all the remaining ones
<Mike Edwards>
Yes you do
<Mike Edwards>
- use constraining type
<Mike Edwards>
see the words in the Assembly spec about constraining type
<anish>
yes, indeed u can with a contraining type. let me think about it. hadn't realized that
Sanjay:
This is really an Assembly TC issue.
