On Thursday, August 7, 2003, at 11:17 PM, John C. Dale wrote:
Dain Sundstrom wrote:
On Thursday, August 7, 2003, at 10:20 PM, jcd wrote:
Concentrating efforts on the basic EJB container services might well prove to be a quicker track to a stable, performing implementation that can be extended to accommodate the various persistence approaches at a later time. It is my feeling that this approach could postpone much of the inevitable analysis paralysis that often accompanies 'The Persistence Debate." Meanwhile, back at the ranch, Geronimo is beating his drums to early release victory.
The initial code includes pool, cache, locking, CMT, BMT, a stateless session container, and a bunch or other basic services. We should be able to quickly have support stateless session bean, stateful session beans and bean managed entity beans. What is missing are big things like a TransactionManager, deployment, JMS, and Webservices.
Am I correct in assuming, then, that you already have a solid idea of how you want to implement remoteable components - stub code generation, compilation, packaging, optimization (if necessary)?
I personally do, but it need to be discussed (but not in this thread). I'll post some ideas tomorrow.
Also, you don't make mention of Entity EJBs above. Do you have any design documentation for the stuff you've implemented? A class diagram, sequence diagram, or even a component model would be helpful.
Yes but there on huge post-it note presentation pads. I could take a photograph of them and post them to the net somewhere, but there is not much information to be gained from the drawings. The source code will be committed when James wakes up (UK time), and you can see for yourself.
One thing I like about the freedom to implement BMP is the opportunity to optimize SQL for my database via hints and proprietary keywords. Have you had any success in building something like this in to your CMP implementations?
Giving users full control over the SQL generated from the persistence engine is a key requirement that went into the new CMP design. Most persistence engines let you do this, but you end up having to effectively take total ownership of the mapping layer. I think that letting users override only the specific parts the need tuning is the killer feature.
We have a lot of work to do, don't we :)
It is going to be so much fun.
-dain
/************************* * Dain Sundstrom * Partner * Core Developers Network *************************/
