Composite Materials

The paper I never gave: XML Europe 2003 on ebXML BPSS and BPEL4WS

I occasionally get requests for a paper that I never wrote! I recently wrote the following message in response to yet another request for the non-existant paper:

---

Unfortunately, the paper you mentioned was never written. Due to a clerical error on the part of the XML Europe 2003 organisers, the abstract for a paper I proposed to develop was published on their web site, but in fact the paper was never accepted for presentation. Due to the organiser's lack of interest, I did not write the paper. However, I can share with you some of my observations about the current standards situation (which has changed since 2003). There are my personal opinions, and in no way are meant to represent Sun Microsystems' stance on these issues.

You are not alone in trying to untangle the world of ebXML and BPEL/WSDL/SOAP/WS-*. It is true that ebXML has not found a large following among software providers in America. The fact that Microsoft and IBM are pushing a separate initiative, known to some as the Global XML Architecture (GXA), has lead most of those software providers to conclude that ebXML is the wrong "horse" to bet on. IBM and MS have rather large coat-tails, after all.

This conclusion has not been shared by many software development groups and users outside the USA. Asia-Pacific, and to a lesser degree Europe, have embraced ebXML largely because it is royalty-free, it works, and provides many useful functions (reliable messaging via internet, interoperability, security, UMM alignment). GXA is largely a set of proprietary specs, with no implementations save for the lowest-level components (which are being slowly standardized). The really useful functions of ebXML are not yet available in the GXA world, and won't be for several years. For organisations that don't have the luxury of waiting a few years to see how GXA shapes up, this presents an interesting choice.

The BPSS vs. BPEL debate was largely the result of marketing "wars" related to the ebXML vs. GXA conflict. Comparing BPSS to BPEL is comparing apples to oranges -- they both have different functions in an e-commerce infrastructure. The emerging picture from standards-setting organisations seems to be:

• At the lowest level, BPEL for performing orchestration of message exchanges. This is a low-level addition to WSDL 1.1, which models message exchanges from a single parties point-of-view only.
• Above the BPEL level, a choreography language is need to properly model the "global" message exchange model between multiple parties. This is being developed by the WS-Choreography working group, and is currently named WS-CDL (choreography descripition language). WS-CDL maps to multiple instances of BPEL.
• Above the choreography level is BPSS. This models multi-party business processes using a subset of the UMM. Such processes are mappable to one or more instances of WS-CDL.

This is the developing picture, based on increasing levels of abstraction, as well as a separation of concerns. The main point is that BPEL is in no way a substitute for BPSS, and vice versa.

In addition, the BPSS 2.0 will provide more direct support for WSDL-described services. ebXML CPP/CPA is providing direct support for WSDL as well.

All of these efforts (updated ebXML standards, and the BPEL/WS-CDL/BPSS stack) are aimed at assuring interoperability between the GXA stack and the ebXML set of standards, in places where such interoperability makes sense. Some folks in the industry question whether the GXA authors will allow such interoperability to truly be achieved, but I don't share that concern: customers have a big influence on such things. This is a Good Thing; this will help drive the important qualities of the infrastructure we are developing / standardising.

I hope this has been of some help.

Best regards,

et cetera

Posted at 02:13PM Jul 19, 2004 by rtenhove in General  |