Events for SCA

InfoQ has a short synopsis of the event proposal that was released to add pub-sub style semantics to the SCA assembly model. The genesis of this lies in some ESB related work I had helped lead at Oracle back in the days when we first started to develop the SCA specifications - it took a long time to socialize this to the working group, but I'm very happy to see that that work has borne fruit. This fills an important gap in the specifications - previously, the wire driven semantics forced components to "externalize" pub-sub interactions in a way that was divorced from the SCA semantics - and is broadly applicable to many problem domains including Complex Event Processing, Queue/Pub-Sub style integration, and ESB scenarios.

I remain somewhat distressed over the way the channel abstraction has played out - the original idea divorced much of this from the assembly view and made it an aspect of system configuration. I think the current approach artificially mixes orthogonal aspects of the problem space. Having said that, this is a milestone for the specification and completes the basic assembly model nicely.

By the way, nice write up Boris!

Business Process Management fallacies

I missed this when it first came out a couple of months ago, in case you did as well: Dubray wrote an interesting analysis of problems in BPM from conceptualization to implementation that is worth reading.

I don't agree 100% with the model for linking BPMN models to BPEL, at least not as the only normative approach to consider, but the runtime model is spot on. The only meaningful model for executable business processes is going to require composite models that incorporate more than just rote BPEL mappings: business rules, human task management, and some form of message mediation. If that sounds familiar, you've probably started to look at the Oracle AS 11 SOA suite preview....

And JJ gets why SCA is so central to getting BPM right. We had a brief conversation about this recently, which I think may make it to Infoq over the next couple of weeks.

SOA platforms

Post updated with correct URL! A high level "white paper" on technology trending toward SOA platforms I wrote a while back, published on oracle.com.

JEE done right?

Sun has been trying to figure out what to do keep pace with innovation in the Java space and especially in JEE. The current proposal adds profiles, which I generally like, since it moves JEE forward in one respect: JEE services a la carte and lighter weight application server technologies. My friend Rod Johnson thinks this is a really great thing and I expect to see Spring take full advantage of this trend.

On other fronts, Peter Kriens asks why Sun isn't leveraging OSGI to its full potential. Well, depends what you really mean there: most of his examples are for app server internals, which can use OSGI regardless of how oblivious the JEE specs are to OSGI standards. But still, with the Java module system being proposed, it does look like we may be in for a dose of infrastructure bifurcation. At a minimum, I'd like to see a coherent and comprehensible strategy in this area for Java.

And I am concerned that there is not a coherent plan for SCA integration into J2EE. In fact, I think the statement on using SCA in JEE is a travesty. What does the current proposal for JEE really mean? How will JEE application expose their interrelationships with other services? And will JEE applications be able to bundle SOA components like BPEL processes used for embedded workflows?

My view is that there are still some very fundamental things to be worked out. We can -- in fact we must -- do better.