Introducing the Semantic Name Service
One approach that has been suggested is that of a Semantic Name Service. The Domain Name Service is the component of the Internet that translates human-friendly system names such as www.ebizq.net into router-friendly protocol addresses such as 126.96.36.199. It is globally available, and it is supported by its users with a minimum of central co-ordination or control, which makes it extremely flexible. The Semantic Name Service would translate between the different terms used by different groups of people and their computer applications, and it should be equally available and flexible.
With such a service, systems that need to communicate can resolve their differences of vocabulary. The two order-processing systems mentioned earlier could find out that they mean the same thing by “quantity ordered” and “number ordered”, and that they mean different things by “delivery date”.
Just as Internet domain owners add details of their systems to the Internet DNS, business domain owners in each organization would add details of their business terms to the Semantic Name Service. Those terms would then be available to other business domains within the organization and in other organizations. The process of mapping between their vocabularies would be massively simplified. And, as with the Internet DNS, much of the hardware and software that implemented the Semantic Name Service would be owned and managed by its users.
But, can such a service work?
Central Index Wanted
The Semantic Web can record definitions of terms and how they relate to each other. It can accommodate different vocabularies – even inconsistent ones. As Tim Berners-Lee says, a critical part of the Semantic Web is the way different communities of practice develop independently, bottom up, and then can connect link by link, like patches sewn together at the edges.
However, the Semantic Web does not provide a service that enables the correspondence between any two vocabularies to be established. If someone has described a correspondence then the description can be retrieved. But there are a huge number of vocabularies, and only a few of the many potential correspondences between pairs of them will ever be documented. What is needed is a central index that classifies all the vocabularies and enables them to be related to each other.