Comments on: BPMN 2.0 Update

By: Ebbe

Ebbe — Thu, 11 Feb 2010 02:40:46 +0000

Mr. Silver, could you please comment on the following two questions:

What is the chance that conformance classes do make it into BPMN 2.0? (25%, 50%, 75%)

If they do make it, would they likely be the ones of Shapiro’s presentation? (Or some mixture of your suggestions and XPDL 2.1?)

This would be very useful to know as a tool developer.

Thanks,
Ebbe

By: Mohamed

Mohamed — Tue, 09 Feb 2010 18:05:08 +0000

Sorry Bruce for wasting your time, I thought I have a different point of view.

I?ll stop this argument, however I?ll continue reading your BPM/BPMN blog, as the best source to learn about BPMN.

Thanks,

Mohamed

By: bruce

bruce — Tue, 09 Feb 2010 17:03:32 +0000

Mohamed,
I think this is a pointless argument. The “metamodel” people in OMG would say that the UML is the “real” BPMN 2.0. Most modeling tool vendors and end users would say the xsd is the “real” BPMN 2.0. If you are saying that xsd, what 99% of the world considers THE definition language for xml, is inherently incapable of describing BPMN, I don’t think many people will take you seriously. Your example is that Task in the UML is a subclass of both Activity and MessageFlowNode (i.e. nodes that can send or receive a message), but the xsd substitutionGroup only makes it a subclass of Activity. I think the reason is sharing common child elements and attributes across Activity subclasses. I’m not sure there are any such children or attributes relating to the fact it can send or receive a message. The importance of MessageFlowNode comes in validating a diagram, e.g. are the source and target of message flow legal in the semantic model? You can accomplish such validation easily in xslt, as well as other validation rules that are not described in the UML. So your statement that xsd cannot handle BPMN is not really correct. Better to say that schema validation with xsd does not validate an instance against all of BPMN’s rules, but I doubt that xmi does either.

By: Mohamed

Mohamed — Mon, 08 Feb 2010 05:39:21 +0000

Thanks Bruce for your feedback.
Let?s agree first that the blueprint of BPMN-2.0 metamodel is the set of UML models/diagrams that come with latest spec. These models define BPMN-2.0 elements, their attributes and the relationships among the elements.

As these models are abstract graphical notations, it does not matter how you represent/serialize them, as long as the representation is correct.

As you said, OMG published two formats; xsd and xmi. My main point is; xsd representation is messed up, it?s not in sync with the metamodel.

According to latest BPMN 2.0 semantic model ?Semantic.xsd? by OMG, there no one single multiple-inheritance relationship is represented in that schema. The reason is simple; xsd does not support multiple-inheritance even with substitutionGroup. The more bad news is; many single inheritance/subclassing relationships (in xsd) are incorrect, sorry to state that, but my reference here is the metamodel (UML diagrams), the blueprint.

Again, as Robert pointed out in his presentation about the struggle to come up with a robust xsd for the new spec, I think this struggle will continue for a while until you find another solution, xmi.

Mohamed

By: bruce

bruce — Sun, 07 Feb 2010 22:59:51 +0000

Mohamed,
This is a long-debated subject. xmi is part of OMG’s MOF/MDA framework and obviously it has its adherents in OMG. To me it is a “proprietary” format because it doesn’t work in real xml tools. Those tools use xsd, the basis of all other process modeling languages, process execution language, SOA, etc etc. The focus on xmi and inability to produce a proper xsd was one reason (not the only one) that BPDM was thrown out in favor of IBM, Oracle, and SAP’s alternative for BPMN 2.0. This is just my opinion, but you ask. Actually BPMN 2.0 publishes its standard both as xmi and xsd. The subclassing behavior is expressed in xsd as substitutionGroups. It’s not exactly exactly the same way UML does it, but it adequately captures the subclass relationship. The BPMN 2.0 semantic model uses this extensively. XPDL does not, and I don’t think it loses a thing, but that’s also just my opinion. I think the reason Robert and I didn’t mention the xmi version is that in our work, which focuses on BPMN tool interoperability, xmi is irrelevant. That’s an xml/xsd world not a UML/xmi world. If you want to use the xmi variant in your work, you are free to do so.
–Bruce

By: Mohamed

Mohamed — Sun, 07 Feb 2010 21:02:31 +0000

Hi,
As I read your update and listened to Shapiro?s presentation, I was surprised that neither of you mentioned the XMI representation of BPMN metamodel, and how it could help solve interoperability and portability of BPDs among applications and tools, respectively.

BPMN-2.0 UML Metamodel is an object representation, but XML is not object-oriented and is not good in representing object models. So, struggling to build XSD for BPMN-2.0 metamodel (as Shapiro pointed out in his presentation) is a normal outcome of trying to tweak a complex object model into XSD. Even if it works, the result will be a bunch of arbitrary XSDs for the same metamodel, since each is a deviation from the standard.

For example, multiple-inheritance is an important feature of object-oriented, and it takes place quietly in BPMN-2.0 metamodel. For example, ?Task? is a subclass of ?Activity? & ?MessageFlowNode? classes. I checked ?bpmnxpdl_39.xsd? schema of Dec 30, 2009, by Shapiro, and it does not seem to have this relationship at all. In general, XSD does not support multiple-inheritance, and I?m not sure how XSD would overcome such a problem.

Another issue is related to cross-file linking, which is represented as associations across the packages in BPMN-2.0 metamodel. Without XPointer and XLink, there is no feasible solution; however, using them is very expensive solution in terms of complexity and maintenance.

I think the solution is already delivered by OMG, as XMI representation of metamodel. It?s a standard built upon XML, and more importantly, it?s a native object-oriented solution made to serialize complex metamodels (as BPMN-2.0) in a simple, straight-forward way while reserving all semantics.

You both are well-respected leaders of BPMN community. I would appreciate any feedback as to why you are only focusing on XSD and overlooking the XMI solution. I feel I may be missing something.

Thanks,

Mohamed

By: Column 2 : links for 2010-02-04

Column 2 : links for 2010-02-04 — Thu, 04 Feb 2010 19:02:11 +0000

[...] BPMN 2.0 Update « BPMS Watch Bruce Silver's comments on yesterday's BPMN 2.0 Update webinar with Robert Shapiro (which I covered in a previous post). He has some great comments on interoperability, especially how the subclasses help with that in spite of OMG's avoidance of any tool interoperability compliance tests. (tags: bpa bpmn) Posted by Sandy Kemsley on Thursday, February 4, 2010, at 2:01 pm. Filed under Links. Follow any responses to this post with its comments RSS feed. You can post a comment or trackback from your blog. [...]

By: BPMN 2.0 Update | ananet.ch

BPMN 2.0 Update | ananet.ch — Thu, 04 Feb 2010 12:26:55 +0000

[...] 2.0 Update Easy AdSense by UnrealArtikel von Bruce Silver zum aktuellen Stand von BPMN 2.0: http://www.brsilver.com/2010/0.....-0-update/ February 4th, 2010 | Tags: BPM, BPMN, Business Process Management | Category: [...]

By: uberVU - social comments

uberVU - social comments — Thu, 04 Feb 2010 06:24:25 +0000

Social comments and analytics for this post…

This post was mentioned on Twitter by BPM_news: BPMN 2.0 Update: Today Robert Shapiro of XPDL 2.x fame, also a member of the BPMN 2.0 Finalization Task Force in O… http://bit.ly/a7nGUv...