A description of the changes necessary to implement named data ranges

Hello,

Here is a description of how the named data range extension would work. In
short, we'd introduce a new type of axioms called DatatypeDefinition. These
would allow you to define a datatype as having some built-in value. Then, you
would be able to write something like this:

(1) Declaration( Datatype( a:myDT ) )
(2) DatatypeDefinition( a:myDT DatatypeRestriction( xsd:integer ... ) )

Note that (1) is necessary because without it, axiom (2) alone would invalidate
the typing constraints (it would use a URI that is not properly typed). These
axioms would be mapped into RDF into (3) and (4), respectively:

(3) a:myDT rdf:type rdfs:Datatype
(4) a:myDT a:equivalentClass ...


We would call datatypes occurring in such axioms '''defined'''. To obtain a
logic with favorable computational properties, in OWL 2 DL we'd have the
following conditions:

- If the axiom closure contains a datatype declaration, then the datatype MUST
be in the datatype map or the axiom closure MUST contain a datatype definition
for the datatype.

- A datatype definition axiom MUST NOT define a datatype that is in the datatype
map.

- Datatype definitions MUST be acyclic.

- Datatype restrictions MUST involve only datatypes from the datatype map - that
is, the datatypes defined through datatype definition axioms have no facets.



All these changes would be reflected in the Syntax document. The impact to all
other documents would be quite small:

- Changes to RDF Mapping are minimal and involve mapping the new axiom (into RDF
and back); both changes are minimal.

- Changes to Direct Semantics are minimal and involve defining the semantics of
the new axiom.

- Changes to the XML Syntax are minimal and involve adding a new axiom.

- There are no changes to the RDF-Based Semantics.

Regards,

	Boris

Received on Thursday, 5 March 2009 12:29:26 UTC