Generating EJB3 annotated beans based on POJOs

I am looking for a way to extract EJB3 knowledge from the source code to external build files.

We are currently doing a migration to Maven of an application originally built by Ant. The application contains over a hundred components with (only Stateless, Remote) EJB2.1 services, running in WebSphere 6.1 with Feature Pack EJB 3.0. For every new service or maintenance we don't want developers to be concerned with specific EJB knowledge, so code is developed in POJOs (as much as possible) and XDoclet is used to generate the EJB-sources and then client/proxy code.

Migrating the XDoclet configuration to the XDoclet Maven plugin is a pain. So we are considering to start using EJB3, but for this we need to add annotations to our POJOs. This would make it impossible for an external release team to describe how the services开发者_如何学Go are exposed.

For example: Is there a way to create EJB3 templates which generate the annotations we need?

(Considering that my premises might be incorrect, I'm open to other suggestions)

---

Could you run your own parser across the EJB source and hence generate the information the release team need?

My thought is that the EJB 3 programming model should be sufficiently expressive from a development perspective. If you build your own generator from some other syntax I fear you'll end up just reinventing EJB3 in a different syntax.