Semantics and SOA: Don't Give Up

Although I don't remember when I first heard the term Services Oriented Architecture (SOA), I remember researching Web services around 2000. Back then, an architecture to handle Web services was unnamed, yet understood - at least to a degree. Now it has a name - SOA.

Back then, it seemed clear to me that Web services could provide more than just a way for Web-based applications to operate. With loosely coupled services communicating via standard protocol, while centralized directories allowed these services to describe their APIs, the sky seemed the limit. Reuse, long-chased but never achieved, seemed almost automatic. Platform independence, long-running transactions, and asynchronous processes - it would be like world peace.Unfortunately, it hasn't really panned out yet. You can't wrap your existing applications (or pieces of them) if you don't know where they are or exactly what they do. Application portfolio management is a sort of oxymoron. Orchestrating these newly independent chunks of functionality is an unsolved problem, not to mention workload management. Instead, vendors have jumped in and created a lot of artificial complexity in Munchausen's-by-Proxy play. As a result, there have been a lot of unmet interoperability promises, expensive failures and only modest successes.

So is it time to give up? Absolutely not.

Last week I attended the Semantic Technology Conference in San Jose. After years of searching, I finally saw some ready-for-prime-time solutions. I'm beginning to feel that semantic technology is ready to step up and solve some of these problems. Here is why I'm feeling optimistic:

1. There are commercially available, scalable RDF stores, notably Oracle and AllegroGraph from Franz. One of the drawbacks of semantic technology has been the lack of enterprise scale tools.

2. With RDF stores that can handle a billion or more triples, and a W3C standard query language, SPARQL, it's actually possible to use large-scale ontologies.

3. The software community has begun to realize that the solution is in the data - SOA won't work without semantic technology, and semantic technology won't work without a different approach to data management - ontology, not E-R or OO [entity relationship or object oriented].

Ironically, Oracle's RDF store is actually embedded within the relational database, but that's the entire idea. At a level of detail, RDF data is just relational. But the emergent qualities of RDF appear when it is treated like more than a pile of records. Web 3.0 is all based on ontology because the W3C realized that the problems of description, interoperation and orchestration can only be solved with approaches than can reason on their own, not by endlessly configuring and scripting things to appear that way.Although I don't remember when I first heard the term Services Oriented Architecture (SOA), I remember researching Web services around 2000. Back then... the sky seemed the limit... Platform independence, long-running transactions, and asynchronous processes - it would be like world peace. Unfortunately, it hasn't really panned out yet.