WebServices dot org

SOA Phase 2 – From the Horse-drawn Carriage to the Car

Model-T vs. Old Paint

I wasn’t around back then, but I’m guessing that Henry Ford spent a lot of breath in the early days trying to convince people that the new horseless carriage was safe enough and reliable enough to use in addition to, or even in place of, the horse-drawn carriage – the proven technology at the time. I believe this example perfectly illustrates the current state of SOA. To clarify my point, let’s visit a more modern example.

Client/Mainframe

When client/server came out, people were reluctant to move from the mainframe, and in large measure didn’t. After a period of years, we saw the benefits and started to add it to what we already had.

Client/server 1.0 was basically putting a GUI in front of an existing mainframe application. This didn’t achieve all of the goals of client/server, but it was a start, and it gave us enough experience and comfort to implement real client server architectures when they were ready.

The above example illustrates a typical path for a new software architecture to follow:

• First, get comfortable with the concept

• Then, add a little bit to the existing technology that you have

• Last, roll out the full architecture when the supporting technologies are ready for prime time

That’s the key, really. Just as most people didn’t buy the new horseless carriage until they were comfortable that it was safe and secure, most IT shops stick with what they know until the new architecture is ready.

So it has been with Service Oriented Architecture (SOA). I believe that we are in Phase 1 of SOA adoption. Like client/server, this has started by adding basic concepts (Web services and related standards) to existing technologies (middleware such as EAI and Messaging systems). This combination of Web services with middleware was a good start, since it allowed people to get comfortable with SOA concepts by using things that they were already familiar with.

SOA Gets Wheels

Just as a GUI in front of a mainframe isn’t really client/server, Web services wrapped around a bus-oriented middleware stack isn’t really SOA . The business agility promised by SOA won’t be realized without making our applications truly service oriented.

Mainframes didn’t go away, and horses didn’t immediately vanish. Likewise, traditional middleware such as EAI, MOM, ESB, etc. will continue to play a role in enterprise software architectures, delivering reliable messaging. But technologies that are designed for SOA have been around for several years and are ready for prime time. For those of you driving toward SOA, the 2nd phase of SOA will provide your "car".