John Halamka, who you may be familiar with as the blogging CIO of Harvard Medical School, also chimes in on the subject today detailing how HMS and CareGroup have used the concept to standardize their information systems from the end-user perspective.

I think that is a tremendously overlooked advantage that SOA can offer; John is absolutely right about the bewildering forest of user interfaces, passwords, navigational paradigms, and data models that propagate in the average enterprise using commercial off the shelf software (COTS). At the same time, nearly everyone recognizes the cost and deployment advantages inherent in COTS, and so it has and will continue to be a primary solution to a lot of corporate IT problems.

Conventionally, SOA is envisioned to exist primarily on the back-end, tying together COTS packages across common services and protocols, with the advantages presumed to lie primarily in the ability to share that information uniformly and discretely. Halamka has turned this vision on its ear, using the common architecture of XML between the varied COTS packages to tie together a unified frontend developed in-house; it presents all the consistency and uniformity that SOA promises, but oriented primarily at the user. That is a tremendous benefit, and when Halamka is judged by posterity, as he alludes to in the last paragraph, hopefully that's the innovation that they will see.

What I hope they will view more cautiously is Halamka's "right tools for the right task" philosophy of development; I think that IT in general suffers from too much of this and that the heterogeneity it has entailed is the root of many of our problems with excessive complexity and poor performance... too often the "right tools for the right task" looks only at the task at hand and not at the larger requirements of the organization. I don't think you need to be dogmatic, but you do need to have standards, and let's face it, the benefits of using the "right" tool for the task are often incremental at best. These are computers, after all; you can do just about anything with just about any of them, the yardsticks that we in IT use when we measure Microsoft against Linux and such are frequently imperceptible to the end user at the end of the day.

I think perhaps if you are John Halamka (a practicing ER physician and dean of technology as well as CIO) or his ilk then maybe it's worth warning yourself against becoming dogmatic, and trusting to your judgement in selecting the best possible tool for the job and that your other efforts will smooth over these extra complexities. For most of us, I think a little more conservatism and standardization might benefit our users further. But something all of us can take away from the post is the concept of using SOA primarily to deliver the best possible experience to those users, instead of simply viewing it as an architectural convenience for IT.