GSA gets semantic with architecture reference models
The General Services Administration is one agency making strides in trying to find a path to the semantic Web through a pilot to encode Federal Enterprise Architecture reference models.
Normally, FEA reference models are written in plain text. Agencies must compare their lines of business to the models and try to synchronize them as best as possible.
The reference models are functional hierarchies, said Dean Allemang, the chief technology consultant for TopQuadrant Inc. of Beaver Falls, Pa., which helped GSA on the project. They have modeled all the FEA models, except the data model, using Web Ontology Language (OWL).
An XML reference model would let agencies more accurately prove that they are compliant with the models than they can do today, Allemang said.
The advantage of the semantic Web is its ability to use the description logic to help carry out tasks, such as cross-checking and verifying consistency within a large project.
A measurement category in the Performance Reference Model, for instance, references a number of business lines from the Business Reference Model.
'In the [orginal] documents, somebody probably copied and pasted them over,' Allemang said. If the original rules are updated, the BRM will not reflect those changes, unless someone makes the changes manually. What a semantic Web approach could do is just reference the PRM metric from within the BRM. So whenever the BRM is accessed, the document will have an OWL axiom that will check the PRM to find out what those rules are.
'For the end user, it would be completely transparent,' Allemang said. The company has devised a demo deployment, which GSA is looking at adopting more extensively.
Allemang acknowledged that the OWL tools now available are still pretty immature. But 'the models [already] have the soundness so you can read them in and compare them in a logical way.'
Owen Ambur, a co-chairman of the Federal CIO Counsel XML Community of Practice said that doing the FEA models in XML is a good idea. 'FEA models themselves are data and, like any other data, can be most efficiently and effectively shared if rendered in nonproprietary XML format.'