1. Download JBOSS, get latest version, why not? (not knowing that DR on the end means Developer Release, so stay away if you aren’t testing it or coding to new features).
  2. Download quick start, only for version 3, download v 3.2, Find webpage that gives all of the missing bits from the supplied quick start PDF at http://jmvanel.free.fr/jboss3-howto.html (thanks Jean-Marc).
  3. Template example will not deploy, gives a bs error about names in the ejb-xml file.
  4. Google leads me to the JBOSS forum page where I find out that the examples are for 3.0, but that you can get them going if you edit the generated files (erm, and lose the changes every time you rebuild). 3.2 is apparently more EJB compliant so it’s more pedantic about stuff generated by xdoclet. I could probably fix this if I knew enough about Xdoclet (or whatever Ant is feeding it) but don’t have the time.
  5. In the mean time I discover that the CMP2 project (which doesn’t use xdoclet to generate the ejb home and remote interfaces or the deploy file, it’s been hand coded) does work.
  6. Fine says I, let’s use this project. Let’s see if the 4.0 DR will work so that I won’t have to migrate to 4 when it becomes production.
  7. It won’t deploy.
  8. OK OK, I give up, I will stick to the 3.2 release, not use xdoclet (I’m writing my own XML generation stuff from a description of my schema, so I can do the interfaces myself anyway).

All I can say is that the supplied stuff does use JUnit and I think JUnit looks OK (but don’t get me started on testing for testing’s sake or you’ll have to kill me). 2 days down the toilet. Lesson : if stuck go to google straight away, do not pass go, just do it. Now adding to my generation code, got the schema, got the drop schema, now to do the bit that generates the ejb-jar.xml.

Next question : do I want to used Container or Bean managed persistence? Container might insulate me even more from the datbase, but bean will give me far more control.