Xtc20031014

Previous: Xtc20031007 next:Xtc20031021

XtC Views about the EJB 3.0 Specification

The attendees were split roughly evenly between those who described themselves as broadly positive and those broadly sceptical about EJB but both sides contibuted to a very stimulating and useful discussion.

At the end of the evening, we took a vote on the many ideas discussed, to determine which three ideas had the widest support from the group. The three winners, in order of most-votes first, were:

• EJB 3 should make use of the Inversion of Control (aka "Parameterize from Above") pattern. This approach minimizes coupling between components that use each other and coupling between components and their container. It also greatly facilitates unit testing.

• A separate JSR should be launched to provide a test container for beans. This could, for example, provide a hook to call setRollbackOnly to facilitate tests for transactionality.

• EJB 3 should provide a standard (portable) way to generate primary keys

The first point is a significant departure from current EJB practice, but could provide clear benefits, as evidenced by its use in both the Spring Framework and Pico Container projects.

Other ideas that enjoyed significant support were:

• Eliminate RMI "pollution" in remote interfaces
• Provide a standard way to invalidate cached entity beans
• Allow polymorphic finder access across multiple types of CMP entity beans

There is still a long wait before we have EJB 3, but these ideas are a good start. Thanks to all for participating.

Here is the blurb written in advance of the event:

ScottCrawford will be coming along to canvas our ideas on the next version of the EJB specification. The group tasked with writing that spec (EJB 3.0) has recently formed, and is eager to hear views from those who are using EJB (and those who might use it someday) about how it could be improved.

You might want to participate if:

• you want to know what might be in the upcoming spec
• you think EJBs are really great
• you think EJBs stink
• you've got some EJB gripes
• you've got a solution to make them better.

If there are clear areas of general agreement about things that should change, Scott has promised to pass them on to the EJB expert group. For those who do enterprise development it should be an interesting discussion!

Preceded by London AgileSIG starting at 3pm

Who's going to be there:

• MichaelMahemoff
• DuncanPierce
• Tony Albrecht
• BenHogan - who is off to australia soon
• AllanKelly
• AlanGriffiths
• SteveF (for a while)

Topics for discussion:

• EjbThree

Questions from those who can't make it:

(Sorry, I didn't see these before the evening so they weren't discussed then but have given answers below. Please note that the answers are entirely my own and do not represent the expert group's or anyone else's official opinion. -Scott C.)

• Are there plans to kill off EJB-QL or will the committee continue to re-invent SQL?

No, I believe EJB-QL should and will stay alive and well. It has a fundamentally different job to do than SQL, even though I'm sure it will continue to be influenced by it. And the specification always provides backwards compatibility in any case.

• Will a solution be provided for applications that wish to use fine-grained domain models?

Have you looked at the changes to entity beans that came in version 2.0 of the EJB spec? Local interfaces have made a big difference in this area, so I suspect you already have what you are looking for.

Who can't make it:

• AdewaleOshineye (I've got flu)
• JoeWalnes (I'm ill too. Probably caught it from Adewale)

The next evening (Wednesday) is the time of KeithB's UpcomingXpTalk at WBS

Remember: ButchersHookAndCleaver, not Old Bank!