[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
|
[news.eclipse.technology.jwt] Re: [Metamodel] Initial metamodel uploaded
|
Hi Florian,
I think inventing new language is a dead end, and it will not improve
interoperability. Instead, we should focus (exactly as you said) on core
features of the business processes, but then just choose the language that
is most suitable (or even subset of it), and write mapping for other
standards that we consider important.
Probably the standard we will choose will not fulfil all our needs (from
graphical, execution or import/export side), so we should extend the format
(most standards supports extensions) and later try to include this
extensions in the next version of the standard. Things should evolve not be
reinvented.
I had a bit closer look on the AgiPro model. And it is a good, clean design
and it covers most of the areas in business processes I know. Problems are
as usual with details... and when we will fix all of them we probably end up
with something very similar to jPDL or XPDL2 or UML2 or other existing
meta-model.
Some of the details that need to be revisited:
* definition of the performer of the activity
* data types - usuall processes can have private complex data
* data mapping - especially mapping between two complex types
* applications - it just cannot be limited to java
* simulation data - it needs much, much more there in order to support even
simplest simulation
* extensability - how to preserve engine/tool specyfic data in the model? do
we need to do that?
* flow - it looks right, but I am not sure. In BPMN you can have
decision/join or parallel split on any activity. Is that the same here?
Other think more on the technical side, the model seams to include graphical
information (in the way like XPDL does) but it probably should not. It
should be down to the editor what information is needed and how it is
stored. We probably should follow GMF way of doing that.
What is in this model that is not available in the other meta-models?
Wojtek
"Florian Lautenbacher" <lautenbacher@xxxxxxxxxxxxxxxxxxxxxxxxxx> wrote in
message news:1ef76d2444eac946b219e4f0828433b6$1@xxxxxxxxxxxxxxxxxx
> Hi Wojciech,
>
> I don't think that we can avoid interoperability problems with the
> creation of a new language. But the focus of JWT is to get an
> interoperability between several process engines and graphical
> representations. Therefore, at the beginning we should evaluate the
> existing languages such as XPDL, jPDL, WSBPEL, BPMN, BPDM, UML2, etc.
> which are currently used by differenz vendors and engines in order to get
> an overview. Then, we can discuss what aspects are needed in our
> meta-model to cover the most important aspects of these languages.
> I don't say that the meta-model from AgilPro is already perfect! Surely
> not! But, during its creation we analyzed UML, ARIS and BPMN and focused
> on the parts that were most interesting as we thought. I just want to
> avoid that we are simply using for example the XPDL meta-model and then
> can't work with other process engines anymore.
>
> What would be the details from XPDL that are a MUST for our meta-model?
>
> Best regards,
>
> Florian
>
> Wojciech Zurek wrote:
>
>> Why do you think that we can avoid interoperability problems if we invent
>> new language?
>
>> The other thing is that probably all current standards were starting
>> exactly the same - XPDL and BPEL and others they were trying to find out
>> what is really important and really needed for process execution and/or
>> vizualization- and trying to store it somehow so the engines/tools could
>> map their behaviours to it.
>
>> (don't get me wrong, I'm not very against new meta-model, but I just want
>> to discuss it as I am not very convinced either)
>
>> Best regards,
>> Wojciech
>
>> "Florian Lautenbacher" <lautenbacher@xxxxxxxxxxxxxxxxxxxxxxxxxx> wrote in
>> message news:6138b59bb43164046e1b9300d8f8de43$1@xxxxxxxxxxxxxxxxxx
>>> Hi Wojciech,
>>>
>>> you are right, this is just another language / meta-model for describing
>>> business processes. I agree, that JWT should provide some toolbox for
>>> working with the existing workflow models. But therefore we need one
>>> basic meta-model where all important informations for workflow execution
>>> can be stored and all other existing languages can be mapped to. The
>>> AgilPro meta-model is just a starting point for discussion what really
>>> is needed for workflow execution. Surely, we could simply take the XPDL
>>> meta-model or the BPDM or UML or BPEL, etc. but then it will be hard to
>>> get the interoperability between the chosen language and the other ones.
>>> Therefore in our first working step we need to analyze the requirements
>>> we have for a meta-model. If there are many things in AgilPro that can
>>> be removed, okay, but first we need to gather the requirements and then
>>> we can discuss how the "interoperability meta-model" should look like in
>>> my opinion.
>>>
>>> Best regards,
>>>
>>> Florian
>>>
>
>