Screaming Penguin - Source Does Matter

I know this is one of those topics we have talked about in this space before, but this weekend's release of Speedo
brought it back up in my mind. I am currently using Castor for persistence in both the new jTOTSP initiative and at work. Castor is a ORM-ish persistence engine as well as an XML-Mapper/Marshaller system that has been around for quite some time. Many people will line up to tell you, though, that Castor is not compliant with the JDO specification from Sun, and I have to say, so the hell what?
The JDO spec, while a fan-friggin-tastic idea has been languishing on a backwater, ass-end-of-code nowhere in the Sun standards world for years now. This is mostly because Sun got their panties all in a wad over CMP-EEJBs, something that only the most devout engage in at any large degree, and left JDO to whither. In the face of this, Castor has plugged on and have shipped a fast, usable persistence engine with perhaps the best set of XML tools out there. Combine that with the XDoclet castor support, and you have about the best ORM development system I can imagine at present.
Now, I think a good application should be loosely coupled to its persistence engine anyway, which isn't to say ever line of code I have written in the last 2 years around Castor has been beautiful, but that's the way it should be. The promise of EJB 3.0 does excite me, and I can actually see J2EE 1.5 changing my mind about how I do persistence, but frankly looking at all the JDO implementations out there and the state of CMP, I will stick with Castor.