Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
RE: [stp-dev] Telco tomorrow - anyone have agenda points?

Hi,
 
For PolicyEditor it will be:
  • Input: nothing
  • Output: XML file contatining policy in WS-Policy format
Regards,
Andrei.

From: stp-dev-bounces@xxxxxxxxxxx [mailto:stp-dev-bounces@xxxxxxxxxxx] On Behalf Of Vincent Zurczak
Sent: Montag, 19. Januar 2009 15:37
To: STP Dev list
Subject: Re: [stp-dev] Telco tomorrow - anyone have agenda points?

Antoine Toulme a écrit :
For the tutorial, we should divide it into time slots, and we should define the input and outputs of each of them, so that we can plug them together.

For example, BPMN modeler time slot:
  input: nothing
  output: a BPMN diagram that shows how to execute a complex choregraphy of processes. Next time slot can reuse the BPMN diagram to generate something executable out of it, we need to agree on that.
I agree with Antoine.

My opinion on SCA.
  • Input : nothing or an SCA diagram.
  • Output : a complete SCA application (assembly and implementations).
The time span depends on the content of the application.
For the runtime, it all depends on what we want to show and how we will call the created application once it is deployed. It can go from 10 to 15 minutes per platform. The interest of the server view is that deployment should go fast (less than 5 minutes).


Here are other elements I've been thinkg about. Things that could be a problem if they're not solved, or at least, discussed before February 9th.
  • [ OK - Presentation ] To create something useful for SCA with STP-IM, the original BPMN diagram needs to be simple. I know a simple flow of tasks works well. Which does not demonstrate all the capabilities of the BPMN editor.
    • One possibility would be to create a complex diagram, to demonstrate the BPMN editor features. And create a new one that could be used in the next step.
  • [ Can Be Fixed - SCA tools ] Transforming a BPMN diagram through STP-IM, when the BPMN is a simple flow of tasks, just creates a composite with one SCA component and one reference for every BPMN task.
    • The process (the flow) is represented not by the composite but by the component.
    • The references implicitely mean the BPMN tasks match existing services. In a simple SOA use case, we might assume these existing services are web services.
    • And at the moment, it is very hard to use web-services in SCA composites. We need to transform WSDL interfaces (got from the web service) to Java interfaces (for the component implementations). There are several existing wsdl2java libraries. Axis, CXF... In WTP, they must have something too. I'm currently investigating this...
    • If you want a summary of the 3 last points, I'm just saying it is difficult and painful to use web services in SCA applications with the SCA tooling. At least, for the moment. I hope to propose a solution for that in the SCA tooling before the end of the month.
    • Another solution is to delete references and implement the tasks by SCA components. We would not use any existing services in the demo. Which is not really SOA, don't you think?
  • [ Unknown - Links between Tools ] Another issue, but maybe Stephane may help about this, is that I (personally) don't know how we could use policy editors with SCA.
    • There are policies in SCA (there are also intents, which are abstract policies).
    • I've read enough about policy editors to understand the way it works.
    • But I've not been far enough in the SCA specification to work with SCA policies.
    • Besides, we don't know how all the platforms handle them. Tuscany does. PEtALS will skip it. I don't know for SwordFish... In any case, it will be very difficult to show them in application at the runtime.
    • There are some work on it.
  • [ Might Be Fixed - SCA Tools ] Last but not least, support of runtime is not yet in STP... I've almost finished PEtALS... but it's not in STP (I should submit it in STP this summer, but it won't be in for Galileo). And nothing is done for Tuscany.
    • In fact, there is something about Tuscany in the archived component SC. But it needs to be refactored and there are dependencies we need to remove to integrate it in the SCA tools.
    • And the way it was done is very similar to Tomcat servers. Which is good and complete. But there may be a more simple approach, using serverdef files (like WebSphere did in WTP). That's also the approach I took for PEtALS. There is much less code to maintain and it is more powerful to add new servers. It would be more convenient to update since Tuscany makes several releases a year.

>From my point of view, there are a lot of things to do or to clarify. It can be for March. I'm not sure it can be for February 9th.
I don't see any problem for everyone to show its tools. But the links between the tools will be hard to show.
Most of all, throught the creation of an "SOA" application. Or at least, in the slides, since they are due soon.

 
Thanks for reading me,

                                       Vincent.

-- 
Vincent Zurczak
EBM WebSourcing
+33 (0) 4 38 12 16 77

Back to the top