[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [stp-dev] Re: [stp-newsgroup] Open letter to STP regarding BPEL and other verticals, from the Eclipse BPEL Designer team

Hi Kevin, Daniel,

B2J was not created initially by the STP project. It originated from TPTP as the 'choreography' component intended to choreograph tests and couple various aspects of the project together using web services.

As such, B2J is in fact quite far on in terms of implementation and the concept of building on model X instead of building on model Y is in fact more like replacing model Y with model X.

However, even this is not entirely correct because B2J does not in fact define it's own BPEL model - it does not create a traversable model of the BPEL at all. It operates directly on the XML file and translates it directly into Java. I would not want to substitute this behaviour with an EMF BPEL model because I this would reduce the portability of the engine (at the moment the entire translation and execution process can run outside of Eclipse in a standard JVM with no additional libraries).

As such I don't see that there is significant overlap between the BPEL project and B2J. If the BPEL designer project can produce valid BPEL 2.0 files then the B2J subproject will be able to use them. BPEL 2.0 is a well defined point of interoperability for the two and I do not see any need for B2J to be any more tightly coupled with the BPEL designer project than through that point of interoperability. The question then is whether to (again) go through a disruptive re-homing process for (as far as I can see) no real benefit.

I would also say that B2J is not the vertical that you cast it as. B2J is a translation framework for vendors to incorporate BPEL process execution into their own engines, servers or applications. That the project itself provides two exemplary implementations of this framework does not make it a vertical.


Daniel Berg wrote:

Hi Kevin,

I am forwarding your note to the stp-dev list as well.

I want to thank Kevin for posting this note to the group. This is an important point that we need to close on quickly. I completely agree with Kevin that STP is a platform for other projects to extend their contributions. We will have some contributions to show how it is done but when there is clearly another project within Eclipse that defines the implementation type we (STP) should not try to declare the contribution to the platform for this type (e..g, BPEL).

I highly suggest that we take Kevin's proposal and run with it.


*Kevin McGuire <kevin_mcguire@xxxxxxxxxx>*
Sent by: stp-newsgroup-bounces@xxxxxxxxxxx

03/09/2006 03:29 PM
Please respond to
"Gateway between eclipse.stp and stp-newsgroup"


[stp-newsgroup] Open letter to STP regarding BPEL and other verticals, from the Eclipse BPEL Designer team

Hi folks,

We (Eclipse BPEL Designer) are somewhat confused about the STP strategy
around BPEL.

Our understanding is that the STP project's goal is to provide an SOA
framework which verticals can plug into.

So why is STP doing java gen of BPEL models and BPMN tooling?  These are

This approach competes with the Eclipse BPEL open source project
(http://www.eclipse.org/bpel/, proposal at
http://www.eclipse.org/proposals/bpel-designer/).  This situation is not
in the best interest of the Eclipse community because we are splitting
our efforts across two models, two UIs, etc.

---Why write your own BPEL component?---

We can understand that there isn't much point having an SOA platform
without some way of choreographing the services.  Given that STP
predates the BPEL project, we could understand why rolling your own BPEL
model and editor seemed at the time like a necessary step.  However, we
do now have a BPEL open source project, and we'd be happy to see its
integration into an STP environment.

BPEL aside, we think there is a larger concern with STP rolling its own
choreography and service components.  There are a number of programming
languages and metaphors which can be used for the choreography of
services. For example, if someone were to come along with a different
open source choreography service, say based on state machines, one would
not expect it to be hosted as subproject of STP.  If you include BPEL,
you should include other services as well, yet clearly STP can't become
"the place for all interesting web services".   Doing so weakens the
notion of STP as a platform.

STP should be exactly what the "P"  stands for -- a platform.  It should
not be in the business of providing vertical applications which will by
their nature compete with other efforts either underway or future. A
clear delineation is required between framework with *supporting*
tooling and vertical applications in order to create a healthy ecosystem
for the development of innovative service choreography components.

Instead, it seems our efforts would be better spent on the integration
glue for the BPEL open source component to live in an STP environment.

---Java Gen of BPEL?---

In order to support Java gen of BPEL, you will need a model.  The
Eclipse open source BPEL project already has a product quality model,
one which has been vetted through the rigors of being shipped in
WebSphere Integration Developer 6.0.  To not use that means you will
need to develop and support your own model (with presumably BPMN tooling
as the visual language).  As we already have a model, that seems a waste
of effort.

Instead, wouldn't it be a better use of our collective resources to Java
gen the eclipse BPEL model?  We would be very open to this and would be
happy to host this effort.

--- Why BPMN? ---

There is nothing I can see from the STP charter that motivates BPMN as
the preferred visual expression of BPEL processes.  BPMN just happens to
be one way of visualizing BPEL flows.  This visualization seems
orthogonal to STP as a platform.


I see there has been discussion on this general topic in the thread:
but the answer to me was not clear (the answer seeming to be, "Yes its
ok we have our own BPEL but sure others can join as subprojects").
There is at its heart here a very basic question of whether STP should
be creating its own vertical components such as BPEL engines/editors.

The counter proposal on the table is one of combined forces, where:
1) The Eclipse BPEL component becomes a client of STP platform APIs, and
2) The B2J work be written against the Eclipse BPEL model.

We believe this approach is the most beneficial to both projects and
more importantly the community at large. We would be happy to host the
STP extension effort for our BPEL engine.

Kevin McGuire (IBM), Eclipse BPEL Designer co-lead
Michal Chmielewski (Oracle), Eclipse BPEL Designer co-lead
stp-newsgroup mailing list


stp-dev mailing list

Antony Miguel
Scapa Technologies
+44 131 550 1766