David Chappell


Get the Feed! Subscribe

Standardizing the Right Thing: BPMN or BPEL?  
# Thursday, June 29, 2006
Creating portable process logic is certainly a good thing. Today, most of the focus for doing this is on BPEL, an executable language for describing processes. As I've argued before, BPEL currently falls short of this goal. Yet even if it didn’t, is BPEL really the right thing to standardize?

Think about the goals we’re trying to achieve. Portability of process logic, the ability to move implementations from one platform to another, is certainly one of them. But portability of people, allowing us to move our skills from one process design tool to another, is also important. BPEL can potentially help with the first goal, but it’s not appropriate for the second; the majority of people who create processes will never work directly in this complex XML-based language. Instead, BPEL is typically generated from some graphical process description. The emerging standard for specifying processes graphically is the Business Process Modeling Notation (BPMN). If BPMN becomes widely supported, as seems possible, it will make people’s process design skills portable.

So wouldn’t it also make sense to use BPMN as the foundation for making the process logic itself portable?The most recent version of the XML Process Definition Language (XPDL), created by the Workflow Management Coalition, defines a way to do this. By specifying a standard XML file format for representing serialized BPMN diagrams, XPDL 2.0 provides a way to move process logic between different environments. Assuming that we can express everything we need to say about an executable process in BPMN--something that's at least a small leap of faith—why should anyone care about BPEL? BPMN and the standard representation provided by XPDL 2.0 can achieve both portability of people’s skills and portability of the process logic those people create. Who cares what language is ultimately used to execute that logic?

There are some challenges in making this a reality, of course. BPMN isn’t yet fully supported, especially by the major vendors, nor is the XPDL 2.0 serialization of BPMN widely used today. Also, although it was created elsewhere, BPMN is now owned by the Object Management Group (OMG). Producing straightforward, usable standards in a timely manner hasn’t been OMG’s forte, so I can’t help worrying a bit here. Also, OMG is hard at work creating a Business Process Definition Metamodel (BPDM) that, among other things, is meant to provide an alternative to XPDL for conveying serialized BPMN diagrams.

Given all of the vendors and standards organizations involved, the most likely outcome is probably a mishmash of partially supported standards, with users left paying the price. Still, the goals are clear: portability of both people’s process design skills and of the process logic they create. Focusing on the graphical process definition notation can address both of these areas. Fully standardizing BPMN, complete with a standard file format, seems likely to be a better solution than standardizing just an executable language such as BPEL.

5 comments :: Post a Comment



David, I like your SOA/BPM blog entries. Makes me alert ;)

If I could draw Java, I would have.

If I could execute UML, I would have.

With BPEL I can both draw and execute.

I've always believed the real value in BPEL is in the abstract. BPEL has nothing to do with drawing or modeling, despite the previous comment. To me the BPEL spec is a platform you build upon - conceptually similar to WS-Security.

I would be interested in your opinion on how SCA can contribute to the portability of BPEL. Since BPEL is being defined as an implementation type for service components it seems that this could foster portability as the functionality not covered by BPEL (that you discussed in "The Case Against BPEL") can be implemented through references and wires.

This is a really interesting comment--thanks. We don't yet know too much about the details of SCA's support for BPEL. Still, it certainly is plausible that just specifying this in a multi-vendor way will nail down some of BPEL's loose ends.

At least one big issue, support for human workflow, would seem to be outside SCA's scope, so I'd be surprised if it helps here. And trying to combine the semantics of BPEL with the semantics of SCA (simple as they are) might well present a few challenges. Nonetheless, I think your overall point is quite likely to be correct: defining a BPEL implementation type for SCA should help make BPEL more portable. Whether the result is portable enough to overcome the things that worry me remains an open question, but the potential for progress appears real.

I've always seen BPEL as a low level "machine interoperability" standard whereas BPMN is more abstract and accessible to less technical, business orientated users.

However, in choosing a standard approach to process modeling, it is important to consider "who" is going to be doing the modeling.

While software developers may well be comfortable with BPEL, business architects (IMHO) certainly are not.

Conversely, BPMN excels in the professional consultant / business analyst arena. The widespread adoption of BPMN is attractive and attractive proposition for this users as it creates a standard approach and a portable skill set.

However, there is another class of business user who wish to construct and execute business processes but who find both specifications intellectually out of reach. These people do not wish to rely on expensive software developer or consultant resources to automate their processes, but would rather get on and "do it themselves".

If you believe that this is a realistic aspiration, then challenge for the process modeling industry is to find a modeling paradigm sufficiently simple, abstract and intuitive to allow these users to express their process requirements, combined with a toolset that is robust enough to allow the processes to be created, validated, executed and managed with ease.

Post a Comment

<< Home