From Bruce Snyder:

Brenda Michelson has made available an awesome BPEL Primer from her research into the WS-BPEL technology. Brenda has a wonderful knack for cutting through a complex spec and applying business meaning to a technology. And BPEL is a difficult one to navigate because it really stands to change the face of business logic in a web services world. Thanks for another great guide, Brenda!

Considering how vague BPEL is to so many people, additional guides containing information are always welcome.

Yeah, BPEL has certain appeal and kind of making sense theoretically but when tried practically… let say it is different story that somehow reminds me Entity Beans saga. It can work, but implementations of the technology are far from being good, they have some very annoying (I would even say showstopper) quirks so the benefits are questionable. See for example this thread Personally I think that Distributed Event-Based Architecture can deliver much more robust solutions ( ORACLEs Fusion architecture white paper is a very good read too).

Yeah, BPEL has certain appeal and kind of making sense theoretically but when tried practically… let say it is different story that somehow reminds me Entity Beans saga.
It can work, but implementations of the technology are far from being good, they have some very annoying (I would even say showstopper) quirks so the benefits are questionable. See for example this thread

Personally I think that Distributed Event-Based Architecture can deliver much more robust solutions ( ORACLEs Fusion architecture white paper is a very good read too).

The DEBA is very interesting, just going through it now. Pretty straight forward architecture and very clean indeed. One issue I see is this model only replaces a subset of a BPEL engine functionality. BPEL, if used as a service invocation state machine, can definitely benefit from simplification. I think the greatest benefit to BPEL is it's standadization of modeling business processes within workflow based applications. You can then move your business processes throughout BPEL implementations as well as communicate them to participating applications without much work. BPEL honestly has many shortcoming to human interaction model, which I believe 2.0 is addressing, though still not completely. Most BPEL vendors have a propriatary work around for these issues, though creating a impedence mismatch between their proposed schema and the BPEL standard. I think that's why BPEL is not being adapter as rapidly as it should be. Other workflow standards address issues very nicely, but then lack web service integration. They mainly don't lack them, rather it's undefined and implementation dependent. One of the main issues with BPEL is reusability of processes. Currently the standard does not allow you to nest and reuse predefined business sub processes that could be used throught the application domain. Either way, I think jboss BPM solution and BPML address these shortcomings, but again in a non-standard way. Ilya

One of the main issues with BPEL is reusability of processes. Currently the standard does not allow you to nest and reuse predefined business sub processes that could be used throught the application domain.

BPEL doesn't actually need a special structure for dealing with sub processes. A BPEL process is acting like a regular Web Service, so you can use the simple invoke-tag to call a BPEL sub process. Kind Regards, Gregory

BPEL is blocks of executions nodes that are put in sequence. It is impossible to jump out of that block. This limitation makes BPEL suitable for orchestrating a service but not for representing a business process, for that something like XPDL should be used. In XPDL you define nodes and transition and does not put any limitation to what node the process can go to.

BPEL is blocks of executions nodes that are put in sequence.

That's not true. There is a sequence and a flow definition. Flow defines a concurrent processes. Ilya

Yes it seems like I have missed something but it is not sequence and flow but the control link. Control links seems to be a "rescue plank" when the structured notion cannot describe what you want. I still think that the idea that BPEL should be structured for human readability is a bad idea. Use a visual tool to make it readable.

BPEL is blocks of executions nodes that are put in sequence. It is impossible to jump out of that block. This limitation makes BPEL suitable for orchestrating a service but not for representing a business process, ...

Let us also not forget the Event handling capabilities of BPEL. Kind Regards, Gregory

very good and informative document. thanks

BPEL is nice as a language for modeliing busines processes. But I don't see much value beyond that for real world problems, until it supports state managment, declarative transaction demarcation and security, among other things.