A reader, Farrukhnajmi, saw my recent post on IBM's bypass of the Universal Discovery, Description and Integration spec (UDDI) and posited this question: If we need a new registry spec, why not look to ebXML?
"I agree with IBM's assessment that UDDI is inadequate for a SOA Registry and Repository. Most importantly, it does not define a repository which is essential to SOA governance. However, I am puzzled why IBM feels that there is a need for a 'new' SOA Registry and Repository standard since OASIS ebXML Registry and Repository provides such a standard already."
The reader states that the ebXML Registry and Repository standard is in fact an ISO standard (ISO 15000 parts 3 and 4). "This is specially puzzling since IBM has already found the ebXML Registry and Repository standard good enough to form the basis of their HL7 ESB offering which they built using the open source freebXML Registry project."
Perhaps for IBM, the huge company it is, it's a matter of one hand not knowing what the other has been doing, the reader speculates. He urges IBM to work with the OASIS ebXML Technical Committee to address any issues that may remain with ebXML Registry and Repository.
ebXML has been around almost a decade, launched in the thick of the e-commerce revolution (as was UDDI). There is still plenty of activity around the spec. A couple of months ago, the OASIS ebXML Joint Committee has created a new XML.org community site. (Ed Dodds even observes that IBM and Microsoft are interacting with the site as well.) Back in 2005, Gartner speculated that there would eventually be a merger between UDDI and ebXML Registry.
A survey of 1,000 companies I conducted with WebServices.Org found that the percentage of companies using ebXML range from 1% of SOA beginners to 18% of advanced SOA deployers. So ebXML does have some traction, though it lags well behind UDDI.
David R.R. Webber recently reiterated the widespread feeling out there that ebXML never received its due from vendors and analysts, who continue to push UDDI and proprietary registry solutions.
Readers -- does ebXML present the alternative SOA registry spec that IBM is urging? Or do we need to move on to something new?