Nothing is more enlightening for a technologist than to observe development in progress. We're faced constantly with a bewildering array of choices and tools. We see specifications on paper that then become something completely different when we actually get to see them implemented in actual software that we then configure to meet our needs, or at least we hope.

I've been spending some time working with a team doing an SOA proof-of-concept test and it's reminded me of what an open book the world of SOA is, and how few pages have really been written in it. The migration of ideas to specifications, and then their transformation within software is a strange process.

Given that SOA has so many optional parts, it's not hard to understand how difficult it is for a vendor to put together a product that actually guides developers in the development process. In what may be the biggest irony of SOA, the technology that we use to enable interoperability is really a set of standalone software, distinct and separate from one another.

If you think about it, there is a logical progression of development for SOA, but because so much of SOA is about enabling communications with existing software rather than creating new services from scratch, there is no one typical development path. This is unfortunate, because the current situation is very similar to a least-common denominator approach, one where each aspect of development is distinct and isolated. You have one console for creating UDDI registry entries, another tool for creating WSDL and other documents, yet another tool for the actual coding of a service, and still another, different place for defining security entitlements. None of which are aware of one another. This makes development a fragmented, disjointed process.

Some may argue that it has to be this way for a toolset to support the broadest range of capabilities. I would agree, but I also think it's possible to create an SOA-focused development tool in the same way that folks like Borland created a Java editor that understood the environments in which it was used. In the same way that code editors today can understand the differences between BEA WebLogic and IBM WebSphere, there is a need for a development environment that understands the various standards as well as the concrete implementations of those standards and how to interface with them to make a development process seamless.

I am well aware this is not as trivial as it sounds. Just keeping an environment in synch with the various levels of specifications is not trivial. Supporting the latest is never enough - think about what would happen if the actual deployment environment is behind in revisions and needs a previous version. Now add to that differing implementations of standards by various vendors and you can begin to imagine the scope and depth of this problem. A good number of vendors have shied away from even contemplating a solution to the issue, preferring to believe there is no solution.

That's a problem, and an opportunity. SOA is too complex to be implemented piecemeal by cobbling together a set of tools. There is a strong need for a product to manage the complexity and variety of the process in a structured fashion. While XML editors such as XML Spy are very good at what they do, what's really needed is a more structured approach to creating services that removes the need to edit XML at all in favor of a more integrated approach that allows the developer to see into the whole process. Simple services are easy enough, but once we start to build complex, composite services that use things like WS Transactions or WS Orchestration, there needs to be a holistic view of the entire process, including the documents and descriptions that go along with service deployment.

This issue focuses on development tools, techniques and practices. We'll show you how to do SOA now, and let you think about how it should be done better in the future.