RE: Question about datatype maps

Hi Alan!

Alan Ruttenberg wrote:

>The OWL specification has a datatype map as a parameter. It's unclear
>to me what are allowable extensions in the realm of datatypes, and
>whether such extensions would be desirable from an interoperability
>point of view. Now that we have a wide range of datatypes, do we need
>to support extensibility here? Our experience OWL 1, OWL 2, that
>choice and semantics of datatypes are not a slam-dunk obvious choice,
>raising questions about whether sanctioned extensions to OWL in this
>dimension would be beneficial or cause more trouble than they are
>worth. Of course nothing would prevent unsanctioned extensions - my
>question here is of what we should encourage.

I simply would not say anything. Of course, we cannot forbid that
implementers support additional datatypes. And custom datatypes, such as
complex numbers, will make sense and add value in certain domains. Interop
is then given by the OWL 2 spec: As long as an ontology only uses those
datatypes, which are defined in the OWL 2 spec, no problems will arise with
reasoners that /at least/ support all these OWL 2 datatypes [1]. If,
however, an ontology uses additional datatypes, all bets are off; but why
should we care as a WG? That's then in the responsibility of those who
create and use such ontologies (they have to settle on a certain reasoner,

Btw, the RDF Semantics document introduced datatype maps years ago, stating
in [2]:
    "RDF provides for the use of externally defined datatypes 
    identified by a particular URI reference. 
    In the interests of generality, RDF imposes minimal 
    conditions on a datatype. It also includes a single 
    built-in datatype rdf:XMLLiteral."

So supporting external datatypes beyond those shipped with the standard was
clearly intended.


[2] <>

Received on Friday, 12 September 2008 11:03:18 UTC