[jbpm-dev] Human-Task module problematic..

Thu Oct 20 09:18:42 EDT 2011

Sure, I can help with packaging the java interface as a cxf war file.  That should be fairly trivial once the straight java interface is complete. Shoot me a message with the interface and I will package it.

Would providing a cxf war with web services and soap over jms as an example help?

Brad

On Oct 20, 2011, at 9:04 AM, "Mauricio Salatino" <salaboy at gmail.com<mailto:salaboy at gmail.com>> wrote:

Hehe.. I agree with you.. but we can't just delete it.. that's why splitting them we can have the core and then add more transport to it.
CXF is fine and we are also using that along with the WS-HT spec interface to expose the module and it works fine.. it just add a lot of boilerplate.
I think that we can play a little bit with the core and start providing different interfaces. I really appreciate kris input on this... because I think that he can give us a good vision about what things we should include and what things we can left out.

My personal roadmap will be:
1) get the core and remove all the unneeded transport layers
2) review the APIs, there are a lot of things missing there and we can decouple the APIs better (users, queries, taskLists, etc)
3) forget about the transport and just get a simple jar with all the functionality.
4) get a detailed plan for integrations (transactions, db configurations, etc)

Most of these things are just copying what is there and break backward compatibility, and if we do that we need to be sure that we will have a synchronization point in the future..

Brad, do you want to help us? what's your github account?

Cheers

On Thu, Oct 20, 2011 at 9:47 AM, Marco Rietveld <mrietvel at redhat.com<mailto:mrietvel at redhat.com>> wrote:
Mauricio,

I have absolutely no problem with first packaging the human-task project in a jar, and then also having it packaged as a web application/web service -- just like Brad described.

I agree with your point about architecture: if you can offer the human-task as a jar, that makes it's inclusion into other systems/applications much easier. And it would be very cool to have it become a WS-HT library that is widely used.

I'll write you/the list probably next week with a suggestion for how we could achieve this: otherwise, feel free to submit a "road-map" yourself!

Lastly, with regards to your pull request/split idea: I don't think that the human-task needs to be split.  I think we just need to delete the transport part. :) We can then refactor the "jbpm logic" part so that  it has a primary interface that the web service or other architectures can interface with.

@Brad: thanks for the suggestion about CXF. Those are exactly the lines I was thinking along.

Thanks,
Marco

10/20/2011 02:21 PM, Mauricio Salatino:
A web application will limit to all the users that wants to add the logic of the Human Task Component embedded inside their own application which I think that is a very good use case. For example, one of the things that I'm pushing forward is the fact that you can create tasks from rules instead of creating tasks just from business processes. If you want to do that having all the logic inside a web application will complicate your application architecture. I'm not against of having a web application that wraps the human task component functionality but we should provide also a simple library that can be embedded in any environment.

I agree with you on the second question, I mean.. that's why I was trying to decouple the logic from the transport in that pull request that breaks apart the current single module into like 8 sub projects. The other thing that it's important to mention is that I believe that the human task component should go outside of the jBPM project to promote the use from different other projects. I would love to see the human task functionality included in JBoss AS for example as a service, but not as a web application :)

Cheers

On Thu, Oct 20, 2011 at 9:16 AM, Marco Rietveld <mrietvel at redhat.com<mailto:mrietvel at redhat.com>> wrote:
Mauricio,

Why don't you think that the module should be a web application?

And why do you want to put the code that implements the server, client and async. nature in the jBPM project??

Thanks,
Marco

10/20/2011 02:09 PM, Mauricio Salatino:
Hi marco,
We have implemented a project that basically provides the bindings for the current module with the WS Interface proposed by WS-HT standard. Doing that we have introduced a lot of boilerplate with the data types proposed in the standard. If you take a look at the new interface that is called TaskService you will see that thats the synchronous interface with the methods that we can expose with WS or JMS(Commands like in drools), but I don't believe that the module should be a Web application. If you talk with kris about this you will find that I have a couple of pending task related with hooking up the task service to the JNDI tree to be able to be used by multiple applications for example in tomcat or jboss.

I'm open to create a project in my github account where we can play around and coordinate that effort, I also have access to a jenkings server to keep it blue. The main problem with the module human task module is that is being used and we cannot break it or change it completely.

-> https://github.com/Salaboy/human-tasks-playground

I've already added you Marco as a collaborator..
I think that the main idea will be to build a proposal to see what we can do and how we can handle the existing users.
Marco what do you think about that? what are the main goals that you want to achieve?
If someone else is interested in this idea please write us back.

Cheers

On Thu, Oct 20, 2011 at 8:58 AM, Marco Rietveld <mrietvel at redhat.com<mailto:mrietvel at redhat.com>> wrote:

Mauricio,

You know the code better than I do, but IMO, the infrastructure code in that module is not okay.

> I really like the idea of having this human task module completely decoupled from the rest of the project

I am all for this. Would you mind hosting it on your github, then?

A lot of the jBPM related code seems to be okay, it's mostly how the module is set up that bothers me.

It seems like it would be a better idea to implement the human-task as a webservice war. That way, we could get rid of the server, client and asynchronous code in the module and concentrate on the jBPM stuff. The JAX-WS standard includes asynchronous webservices, and webservices can be coupled to JMS implementations as well.

For more information on JAX WS asynchronous services, please see the following links:

 *   http://today.java.net/pub/a/today/2006/09/19/asynchronous-jax-ws-web-services.html#asynchronous-computation-with-future-and-callback
 *   http://publib.boulder.ibm.com/infocenter/wasinfo/v7r0/index.jsp?topic=%2Fcom.ibm.websphere.express.doc%2Finfo%2Fexp%2Fae%2Ftwbs_jaxwsclientasync.html
 *   http://biemond.blogspot.com/2011/02/building-asynchronous-web-service-with.html
    *   (not especially about async services, but gives you an idea of what the code looks like)

Thanks,
Marco

10/20/2011 01:27 PM, Mauricio Salatino:
Hi Marco,
My opinion about it is good, I mean, we need to improve the transport layer of it, but the rest of the code is ok. If you take a look at one of my last pull requests, it was related with the fact that we need a synchronous interface to simplify all the interactions and to hide all the boilerplate of async communications.

If you want to simply things to start having more smaller and controlled pieces I suggest you to take a look at this other pull request: https://github.com/droolsjbpm/jbpm/pull/28
Which aims to split the logic that we currently have inside the human task module into several modules with their own dependencies. That will allow us to do internal changes, api changes as well as transport changes in a more manageable way.

I really like the idea of having this human task module completely decoupled from the rest of the project and I'm willing to help as much as I can to improve it because I found that not only jBPM can leverage the power of it.

Let's keep this discussion open so we can gather more requirements from the people that is using it.

Cheers

On Thu, Oct 20, 2011 at 7:19 AM, Marco Rietveld <mrietvel at redhat.com<mailto:mrietvel at redhat.com>> wrote:
Hi guys,

Having looked through the architecture of the Human-Task module in the last month or so, I've become fairly pessimistic about it.

The biggest problem I'm seeing is that the "wheel is reinvented" a couple times -- and the "reinvented wheels" that are present in the Human-Task service will be a pain to maintain/troubleshoot.

The "reinvented wheels" are things like the following:

 1.  The server logic
    *   BaseJMSTaskServer, TaskServerHandler, etc.
 2.  The client logic
    *   TaskClientHandler, the ResponseHandler and all it's children classes)
 3.  The asynchronous/concurrency logic
    *   AbstractBlockingResponseHandler{.waitTillDone(long) } and every class that uses that method (and every class that uses the class that uses that method.. etc.)

 I think my frustration with this can best be expressed by the fact that jBPM is a process engine project -- it's not a server project, it's not a (service) client project, and it certainly isn't a project that supports asynchronous communication. And yet, we're implementing all 3 in the module. :/

Lastly, the human-task module code is the reason that the jbpm builds (on hudson.jboss.org<http://hudson.jboss.org>) have been failing for the last month or so. And the tests are not failing because the tests are wrong: the tests are failing because there's a race condition in the code, and it occurs when you run the human-task code on a 1. heavily loaded server that's experiencing 2. lots of network traffic. Which is what the hudson.jboss.org<http://hudson.jboss.org> is.

I guess I'm wondering what other people's opinions about this are!

Thanks,
Marco

--
jBPM/Drools developer
Utrecht, the Netherlands

--
 - CTO @ http://www.plugtree.com
 - MyJourney @ http://salaboy.wordpress.com
- Co-Founder @ http://www.jugargentina.org
 - Co-Founder @ http://www.jbug.com.ar

 - Salatino "Salaboy" Mauricio -

<ATT00001.c>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/jbpm-dev/attachments/20111020/0d345e3d/attachment-0001.html 