Registry: every SOA needs one

Some reactions to my post yesterday regarding Yankee Group's optimistic projections that 75%-95% of enterprises will have functioning SOAs in place in 2006:

Jorwell cuts right to the chase, noting that if 2006 is the year of SOA, then "2007 will be a busy year, clearing up all the mess left behind by the failed SOA implementations. Didn't your mother tell you never to implement anything that hasn't been properly defined yet?"

Rebecca O'Neill of Capient told me that she feels "Yankee Group is very misguided in their SOA report... It's no question it will be a number of years before you might see a ramp up leading towards critical mass of XML Web service capabilities across organizations."

Capient believes that  there's still a key piece missing to the SOA puzzle -- registries. There has been movement in the industry around UDDI, but this standard is still in the early stages. Registries are essential to helping SOA users, administrators and developers locate services that can be reused across the enterprise.

"It's still an unknown how XML Web service capabilities and publication registries will evolve, mature and ultimately reach critical mass," notes a report from the analyst firm. "A variety of registries currently exist, ranging from the public consortium-led Universal Description, Discovery, and Integration network to stand-alone, prepackaged and vendor-managed UDDI specification-compliant registries and hybrid non-UDDI-compliant registries following public and/or proprietary specification formats. by third-party vendors." Ultimately, Capient believes, SOA and Web services developers will rely on a combination of all types, depending on need.

Part of that mess we're cleaning up in 2007, then, may be trying to figure out what we did with the services we had written in 2006.