As a consultant in Business Improvement and Alignment, I read with horror your reference above to the Dutch Government that has a quarter of a million activities for one process. I think this is impossible unless they have used process notation to describe all of their checklists…

I understand that this is a technical discussion, but I think the biggest problem that most companies face is the imprecise nature of business analysis and the training given to Business Analysts and Process Analysts.

I have personally taken a printout of one “process”, developed by a major consulting company and laid it out on the floor. When I connected all of the pieces of paper, the thing was 5 meters long and activities at the far edge of the process linked back to activities at the start of the process.

No-one could work out what was happening, so no one from the business would sign off on the process. I produced a framework and condensed this into 4 separate one page processes, and the business agreed with the process straight away.

I think you will find that documenting activities at too low a level and failure to establish a process framework are the two biggest mistakes analysts make, and these have a huge impact on the final product. Following close on their heels is inconsistent use of terms, process descriptions, and patterns.

BPMN is great to describe processes to business users, but not to the extent that you can generate BPEL from it. The reason is that in order to get agreement about processes, business users generally should not be exposed to the sheer number of symbols that make up BPMN. The complexity overwhelms them, whereas it is the symbols and the application of their precise meaning that permits them to become BPEL.

And finally, people with the skillsets to take the process from the business and into IT are very rare. Very rare.

-Rob

I’m not sure about that. I think both are needed for making educated business decisions. Also, while designing executable processes requires the skillset of a process analyst, developing BAM dashboards usually does not, therefore I would believe that a business analyst could easily handle both BAM and BI activities.

Best regards
 -Ismael

What is your opinion on data analytics vs. process analytics? You mentioned that BPM should be used by process analysts. Would it make sense to separate process improvement activities (BPM scope) from other business improvement activities (business intelligence scope)?

If you are looking for a forms-driven solution, look at Karomi. It is a workflow tool, but it has most of the features that a BPM tool would have. Web-based process designer, Web-based form designer, and a mini BAM.

-Joe

I’m thinking of an all-encompassing business suite that incorporates not just process and all of its supporting capabilities, but also content, integration, collaboration, security, identity, metadata repository and a host of…

Give Intalio a try, look at how we implemented workflow patterns on top of the BPEL engine, and let me know what you think. I believe that it will clarify a lot of things for you.

Best regards
 -Ismael

The BPEL semantics is good for small and manual problems, but I believe that an automatically-generated BPEL process has no “human semantic” at all. And BPEL doesn’t include important new high-level concepts (except correlation).

Runtime changes… The executing process is a BPEL one, you told it is “computer code” so nobody can touch or modify it. So people modify, in their development environment, the BPMN process and generate the new BPEL. Of course modify one “arrow” in the BPMN process may change completely the resulting BPEL — unstructured vs. structured problem again. Then they deploy the whole new BPEL in the execution platform. And how could you migrate existing processes instances to the new BPEL process? BPEL has no concept of state, so it is hard to imagine how to do that dynamic changes and migrations at runtime.

And What about BPMN2XPDL? Don’t you think it is a more natural mapping — with a fully closed, portable and executable XPDL standart of course. I like BPEL, but not for BPM. I like BPMN, but it allows to model lots of really evil things. XPDL is nice also, but it’s full of non-portable vendor extensions. So I don’t know what to think. I am still playing with my Petri Nets meanwhile…

Thank you for the answers.

The thing is, if you want the ability to execute a very large number of processes, in real-time, leveraging the full semantics offered by BPEL, and letting users dynamically change the definition of processes at runtime, building a BPMN2Java compiler will be an order of magnitude more complex and difficult to maintain than building a BPMN2BPEL compiler plus a BPEL interpreter. If you need proof for this, look at the database world: the Oracle database does not have a SQL2C compiler. It has a native SQL interpreter. For the very same reasons.

I understand that BPEL is a strong standard for composing new services through to Web service orchestation, without unstructured patterns, without standard human interaction, and within a SOAP/WSDL world). But it’s not a BPM language, and it’s not good for machine code either. In fact BPEL is a new programming language, quite similar to any other, but with some nice high level concepts… But is that enough? I think that I can do a BPMN2Java compiler with less effort than a BPMN2BPEL compiler plus a BPEL interpreter. My computer doesn’t care.

My poor english doesn’t let me be as polite as I wish. Thank you in advance.

-Marcos

Because computers will use it, and that’s all that matter at the BPEL level.

First of all, sorry for my bad english.

I have a little question. Business processes are, by nature, unstructured processes (like BPMN is). BPEL is (should be) a structured programming language. Migration from an unstructured to a structured process is a non-trivial task. Indeed, some processes that can be modeled in BPMN with paralelism and arbitrary cycles cannot be translated into BPEL processes.

In cases where mapping from BPMN to BPEL is possible — of course not just the trivial BPEL mappings provided by the BPMN 1.0 specification, the techniques employed are really complex and generate unfriendly BPEL. I believe that BPEL links and event handlers (unstructured BPEL elements) should be used to model many elements of unstructured scenarios. I guess this is how Intalio does it anyway.

At the end of the translation, you’ll never recognize your business process in the resulting BPEL code. So, why choose BPEL as execution language if users/analysts/programmers will never use/touch/understand it?

-Marcos

I am new to BPM, more interested to know about its connection with Knowledge Management. Since I read some few articles on KM-BPM, it would also be helpful if I could get some career opportunity guidance. I am working on a Knowledge Management Implementation.

Regarding your first question, I would say that orchestration and choreography complement each other. Orchestration (BPEL) describes the high-level business process, while Choreography (WS-CDL) can be used to execute low-level transactions, especially when acidity, consistency, isolation and durability are required behaviors, but cannot be implemented using traditional transaction coordination techniques. For the BPM developer, choreography should remain largely invisible though, and it becomes the BPM tool’s responsibility to generate the appropriate BPEL code and use web service interfaces that support the WS-CDL model to execute transactions in a reliable way.

Regarding your second question, I believe that the BPM 2.0 manifesto should not mandate the use of any particular programming language, be it Java or C#, for the implementation of the BPEL runtime component. Instead, it should be the BPM vendor’s responsibility to expose it’s back-end infrastructure in ways that will allow customers to take advantage of it in the most productive way, according the the programming model they use. In Intalio’s case, our connector framework is implemented in Java, supports both the AXIS and JBI models, and allows customers to build their own connectors. This should address your needs I believe.

I am from a J2EE developer background. In fact I lost my virginity to Business Process Modeling only a couple of days ago. So at the moment, chaos reigns supreme in my head.

1. I have done some reading on WS-CDL (Choreography) and BPEL. I also had a look at the Open Source pi4soa. Does BPMN+BPEL (Orchestration) as defined in BPM 2.0 make Choreography redundant?

2. I find it troubling that BPM 2.0 has thrown away the policy of having BPEL generate Java for the Web Services and uses some other (perhaps proprietary) mechanism. For those who work with BPM 2.0, it wouldn’t have really mattered how everything is being executed. But in some cases, when IT people need to plug into the BPM stack, it would have been great help if everything underneath was accessible. I think BPM should not turn away from using open standards at the back-end as well. This would be a major mistake.

As part of my recent exploration of developing strategies for using Web 2.0 in the enterprise, I find that time and again the lowly wiki presents itself as the most likely target for the initial adoption in the enterprise. For one thing, almost everyone…