Relation Registry Outline Specification

• We make a clean separation between the basic backend technology and possible command-line, desktop or web-based frontends. The initial frontend will be command-line based.

• The interface between the RR and other components in the CMDI will be via REST-based HTTP.

• Adding and deleting relations will be via PUT and DELETE operations on XML representations of RDF graph fragments defined by an OWL schema.

• Retrieving of the information from the RR will be via GET operations containing SPARQL queries.

• The relation information will contain the following pieces of information:

• Properties - two meta-data properties over which the relation is being defined. In the initial implementation, these will be opaque strings. Many of the properties will come from nested XML structures. The string could therefore be in a dotted notation defining the path through the XML tree. This might be shared with the query language offered to the end user. Or, perhaps, the string may be an XPATH expression. In any event, the initial implementation of the RR will simply store and retrieve strings without examining their structure.

• Relations - to be taken from an ontology defining the possible space of relationships. Menzo has offered some ideas.

• Context - A string that may represent a user, who maintains personal property mappings, or perhaps groups of users.

To come:

• RROWLSchema?

• RRExamplePUTOperations?

• RRExampleGETOperations?