This latest gosspi on fake posters is funny. Just check out how many alter egos I have. Can anyone beat me?

Ok, first of all. I actually think EJB3 will be a massive improvement compared to previous version. They really seem to be on the right track.

But I'm going to prefer not to do EJB anyway.

Jason Carreira touched upon in it in this blog entry. The reason is simply that the choice of app-server (and thus EJB container) is usually not a technical decision. It is a political/strategic/whatever decision. When I come into a project at a customer all those decisions have already been made and a simple consultant will rarely (never?) be involved in those decisions. Here's your app-server, here's your database. Now, build me this bad-ass system.

This means that if there's a bug somewhere in the app-server you're looking at a 3+ months cycle of nagging on IBM support until they assume responsibility, waiting for them to actually figure out a way to fix the problem and finally persuading the customers production department to apply the fix. And that's assuming you don't hit a brick wall somewhere and it's just not going to happen.

In a situation like that you got one choice: rely as little as possible on the app-server/container, and always expect the worst. We're talking basic servlets here, you'd be lucky if you get a jsp running and don't even dare thinking about servlet-filters, EJBs or anything fancy like that.

With Spring/Hibernate (or whatever is the current vibe) I bring my own lunch. I can build the entire solution as simple servlets. I can run it standalone with Jetty inside my IDE. If there's a bug somewhere I'll just go in and fix it. If there's a nasty deadlock under heavy load in production (or whatever) I can figure it out. In a put-your-trust-in-the-holy-container-provider-situation my ass is grass.

So thank god for lightweight enterprise Java (as it's currently seem to be called)! You've saved my ass so many times.