I just listened to a very excellent podcast from Dr Dobbs Journal about Service Oriented Architectures (SOA) and Object Oriented Programming (OOP). The SOA philosophy is gathering a great deal of steam, especially in places where OOP proved itself to be inadequate. A reader asked the question, does SOA mean an end to OOP?

The very idea makes a lot of UML and Business Objects fanatics VERY nervous...

I believe the podcast touched on all the important issues, and validated what we have been saying at Stellent for almost eight years: OOP is a fine way to create applications that perform specific tasks. However, it is a terrible architecture for creating robust network applications. Every OOP-based architecture for distributed applications (using CORBA, DCOM, or EJB) has failed to deliver on its promises... and made network programming vastly more difficult than it needed to be.

A service-oriented approach, on the other hand, gives the developers and architects an easy way to integrate systems. They are far less complex, faster, more robust, more scalable, more easily maintained, and they are much easier to for designers to conceptualize.

Does this mean an end to OOP? Good heavens, no. We still need OOP -- or at least significant chunks of it -- to create these systems. However, we should stay as far away from OOP as possible when creating interfaces to distributed applications.

It will probably take several more years for SOAs to become common practice across the industry. And until then, people will demand an object-based API to network applications. The number of SOA converts has not yet reached a critical mass, so creating OOP interfaces still has value... if for no other reason than that's what people expect.

I believe SOA will become a no-brainer as soon as somebody solves the distributed authentication problem. Without it, SOAs are only good for intranets, and specific extranets. I have been watching SXIP for about a year now, and I think they are closer than anybody... but only time will tell.