[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [stp-dev] Intermediate Meta-Model Discussion - Update
|
Hi Adrian,
Looking at the hybrid model that you've proposed, I can get the
Service, Owner and ServiceClass entities. Are the pieces to the
right of that for capturing the BPEL / BPMN pieces? I'm guessing
that to be the case, but would just like to make sure.
Exactly, we thought these pieces would be a sort of the bare minimum
to have in the model, things that might map [easily] from BPMN/BPEL
to something like SCA and JBI. An example would be of a process made
up of several steps that invoke services which would map to an SCA
composite having those services inside, or to a JBI assembly.
Obviously, a representation in SCA or JBI of a process does not carry
at all the same semantics as it does in BPMN/BPEL. However, the point
is that we'd carry as much information as possible to these editors
so that we remove work duplication.
Does the Service entity need some more fleshing out? I'm not sure
how the SCA & JBI models which hold info such as bindings maps over
to this, or if even you would like to bring this information into
the model. I know we're aware that there may be some data loss when
going between the models, but a hybrid should capture all the
information that would ultimately be key to execution on a
particular runtime.
I think your question raises an important point: do we put runtime-
specific information in this model (such as binding information) or
do we leave it at a "logical" level, i.e. functional architecture?
Since the purpose of this model is to carry as much information as
possible and needed between editors (rather than being a pure breed
model of any particular kind), I think you might be right, and we
could enrich it with this sort of information. For editors where this
would not be applicable (like BPMN?) we would simply not save/use
these pieces of data.
One thing that takes my interest is I'd like to be able to use two
additional editors not mentioned here, a WSDL editor and the
eclipse JDT Java editor. So when doing Java first for example, I
have a bunch of annotations that define my web service. These would
certainly contain the interface definition, but also potentially
the more physical information about the service such as binding /
transport that it runs on. In addition the WSDL model coming from
the WSDL editor would need to be represented as an interface and
some physical information. So if we're publishing this into the
hybrid model, I think this info would be required.
I agree, similar to above.
I'll stop here for now, my questions might be my lack of
understanding of the model. Perhaps you'd have some time tomorrow
to join our STP planning call and we could look into this in some
more detail?
Yes, I will take part in tomorrow's discussion in the STP planning
call, and my colleague Alain Boulze might join in as well.
Looking forward to talking to you guys!
best regards,
Adrian.