Blogging Roller

That's three JPAs. I've been fighting to get a new Roller back-end running, one that uses the Java Persistence API (JPA) rather than the native Hibernate API that we use now.

Craig Russell and Mitesh Meswani wrote the new JPA back-end and tested it against Toplink/JPA (JPA #1) and the Derby database. I've been refactoring it, updating it for Roller 4.0 and testing it against various other databases and JPA implementations with mixed results.

I tried to run it against MySQL, ran into a bunch of problems and fixed those. Allen switched out Toplink/JPA for Hibernate/JPA (JPA #2), ran into problems and fixed those. I switched out Toplink for Apache OpenJPA-incubating (JPA #3), ran into problems and fixed those. So now we have a back-end implementation that runs against OpenJPA, but requires two small code changes to work against Toplink -- not exactly an ideal situation.

I'm a JPA newbie for sure and still trying to figure this stuff out, but it does seem that switching JPA implementations is not as easy as it ought to be.

In other JPA news...

I didn't realize that the open source version of Toplink that Oracle contributed to Glassfish, Toplink Essentials, is the pee-wee version of the full and still closed source Toplink product. That's why I was somewhat surprised to hear that Oracle is open sourcing and donating the full version of Toplink to the Eclipse foundation.

Two observations about that news. First, this move appears to be a fork that will result in a CDDL licensed version of Toplink Essentials in Glassfish and an EPL version in Eclipse -- I hope I'm wrong about that. And second: apparently IBM will be soon be supporting two versions of JPA: OpenJPA via Geronimo and Toplink/JPA via Eclipse.