As the
Plain Old Java Object (POJO) ascendancy becomes increasingly widespread in the
Java community, it appears likely that one master framework for exerting Inversion of Control will emerge and grab mindshare. If this were possible, enterprise application architecture could be improved enormously, especially if it conveniently allowed developers to avoid the ugly internals of J2EE along the way. Of course, I'm talking about
Spring, that incredibly handy, elegant, and well-done framework for making it easy to build well-designed Java apps. In fact, I've repeatedly predicted in recent months that the forthcoming EJB 3 standard will likely be eclipsed by superior, and easier-to-use, enterprise patterns like the Spring/Hibernate IoC/Persistence combo.
A brand new blog entry earlier today by Java luminary Dion Almaer certainly heads in the same direction and makes some of my points. He says that major vendors have pointedly told him that EJB 3 annotations in Spring are easily source level compatible, so there is no huge need for EJB 3. This means that organizations that don't need full EJB 2.1 backwards compatibility can avoid the sky high costs and byzantine complexity of commercial J2EE containers, all the while reaping the benefit of a clean, highly usable, and oh-so-malleable enterprise programming model. These same organizations are likely to be happy to be free of the shackes J2EE complexity for future development. Instead, if they need heavy-duty distributed transactions or enterprise directory services, they can just pop their EJB 3 source code compliant components out of Spring and put them into a "real" EJB container. This makes the EJB 3 standardization folk's decision to go the annotation route brilliant. Brilliant for not making everyone have to use a J2EE container that is.
Figure 1: The Spring Framework makes a terrific annotation-aware Java container.
In reality, some development shops won't be able to immediately adopt Spring-like frameworks wholesale. But the longer term advantages in terms of improved development productivity, reduced maintenance, and easier management will continue to make an obvious and hard-to-ignore case.
The compelling question in my mind is whether EJB 2.1 backwards compatibility, which a recent The Server Side Symposium poll showed that a whopping 80% didn't care about, is worth holding back a significant new development in the Java community that the .NET community is still trying to deal with: a vastly simplified and improved programming model.
Inversion of Control, annotations, and aspect-oriented programming stand to change the playing field by putting incredible new powers into the hands of everyday programmers as a readily accepted practice. Projects like Spring, which are already known as J2EE without the EJB in many circles, are clearly in a position to deliver this to folks today.
Technorati: Enterprise Architecture, Computers and Internet, Programming