I found an interesting post about the role of semantic data integration in SOA by Ken Rugg. The author's main argument is that an efficient system governance implies an efficient SOA governance. And that this SOA governance can only be achieved if the data that is processed by an SOA architecture has the same meaning everywhere and for everybody. The way the author sees semantics is in the form of a Common Informational Model, not in the sense of reasoning-enabling ontologies. This reminded me of the CIM models by the DMTF, for IT system management, as part of their Web Based Enterprise Management (WBEM) activities. The efforts driven by the DMTF also seek to enable efficient distributed systems governance. Nothing exceptional up until now, apart from the fact that semantic data integration is not easy at all in highly dynamic systems, such as it would be the case once web service compositions will become a reality. No, the most interesting is Progress Data Xtend. Their approach to the question of enabling semantic data integration is using rules. However these rules must have access to all the data exchanges which from my point of view is not possible unless you use a service bus for managing and executing services, because it is a closed system, and controls all data exchanges running upon it. While many people think that SOA can only be implemented through a service bus, this is not true. SOA says nothing about its implementation, and SOA doesn't even tell you that you need web services. One can imagine open systems able to dynamically generate composition of their services, and a new choreography for all their services, once they have discovered a new service. Nevertheless, the idea would make sense if you use one single platform like JBPM from JBOSS to execute your services and are ready to integrate the Data Xtend platform to the rest of the SOA architecture. An interesting approach in all cases.
Marwane El Kharbili.