Despite the PR hyperbole that tends to drown information-technology (IT) users in buzzwords, there is really no such thing as open-source service-oriented architecture (SOA). SOA is a blueprint of one way to build IT infrastructure, one that will most likely dominate the market through about 2020 because it adds the concept of utility-like computing to the re-use and interoperability promised in earlier IT architectures such as client/server (C/S). Open source software (OSS) is a potential SOA building material—the brick and mortar, studs and nails, windows and doors—just as it has been used in C/S computing for years.

OSS is not required to implement an SOA and an SOA design does not need OSS. But they go together nicely. Although it is possible to have a more modular rather than componentized SOA design, the SOA concept really does not make much sense without hundreds or thousands of highly granular components, or services. These fine-grained services need to be decoupled from individual applications (and pieces of infrastructure software) and stored both in services repositories (see illustration). There will be two types: those behind enterprise firewalls and “community repositories” organized by supply chain or other inter-enterprise characteristic. Community repositories do not really exist yet and are expected to emerge in the 2009-2012 timeframe.