RE: BPSS and ISO T68/WG4

Steve Ross Talbot writes:

> I got the minutes from our last ISO WG4 meeting. Included is the

> following:

> Q: How does the choreography compare with that in WS-CDL?

 

> A: The choreographies are not very different, but if you port BPSS 

> into CDL a lot of material is not carried over. However the BPSS 

> choreography was developed empirically from actual projects and is not


> based on Pi Calculus or any other academic body of thought.

> 

 

DaleMoberg>> Both Monica Martin and I responded to this question and we
mentioned (among other things) the following as contributing to BPSS

 

UML and 

UNCEFACT's UMM and

BPMN (extensions were used in diagrams)

Use cases from UCCNet, RosettaNet, Twist, and other Vertical SDOs.

 

 

I mentioned that the UML 2.0 Sequence Diagram with the Combined Fragment
enhancements (for alternatives, options, par, break, and so forth) is
similar in expressive power with respect to choreography. 

 

There was little explicit consideration of pi-calculus in BPSS
development. This comment is not meant to be pejorative of either BPSS
or pi-calculus.

I did indicate my opinion that viewing business collaboration as a
distributed computational process, while possible, was not expressed in
the specification. 

 

> What specifically is lost when xlating from BPSS to WS-CDL? 

 

There has been no specific investigations on expressive powers, but this
comment was probably in the context of mentioning that the business
contract documented in a BPSS M1-level description is not only the
messages exchanged together with the "logic" of the exchange, but also
many quality of service parameters pertaining to the process. I do not
think that these features have been a focus of WS-CDL at least when I
last reviewed it. That would be understandable because these features
pertain to aspects of the business contract that are not primarily
choreographical. The various timers over "stretches" of message exchange
that pertain to "time to perform" or "time to acknowledge" are such
features. Likewise, the parameters pertaining to QOS of reliability or
data confidentiality or data origin authentication are such QOS aspects
of the contract. So these components of the model drop out. A similar
situation would be found in maps to UML or BPDM. The requirements on
BPSS documenting a business process, and the abstraction level enforced,
are distinctive. 

 

I guessed that there were probably no choreographic features that would
necessarily be lost in forming a WS-CDL document. Possibly you would
need to include, in the BPSS document, some information in the (BPSS
optional) OperationMapping elements that say how to connect a
BusinessTransactionActivity and its referenced BusinessDocuments to a WS
portType/interface's Operation and messages. This matter depends on how
"abstract" the connection to a wsdl defined WS can be. I will let Monica
who knows both specifications comment on how easy a translation would be
that did not contain OperationMapping.

 

> 

> What would be lost in xlating from WS-CDL to BPSS?

> 

 

We did not discuss this direction of translation. 

 

I am uncertain how the rest of the remarks pertain to anything we stated
in our review of BPSS. I would note that both UML and UMM (which are
formalisms of course) did provide foundations for constructing BPSS and
its schema.