Re: draft public comment for OWL last call from RIF

On 26 Jan 2009, at 15:41, Chris Welty wrote:

> I'm not current on the OWL spec so I don't know exactly what you  
> mean by "support" here.  RIF does allow for user-defined datatypes.

Ok.

> The RIF BLD conformance clause paramaterizes RIF conformance wrt a  
> set of supported datatypes, however RIF does not provide any  
> particular mechanism for defining them, I guess we've been assuming  
> users would use xsd for that.

Ah. OWL now has built in syntax for that as well as restrictions on  
what "makes sense". These presumably could be more liberal for rif if  
you only have to check instances. But then I don't understand the  
problem with the derived string types. They are expressible in XML  
Schema and it seems a bit odd not to allow the standard names there.

Also, if conformance paremeterizes why not provide a set which is  
"core minimal" and one which is "owl"?

Consider owl:rational. There's a strong prima facie semantic web  
argument for it: On the semantic web one should be able to write 1/3.  
(One should be able, in principle, to write all mathematics, but we  
can compromise on things which we can effectively support.) OTOH,  
it's not clear that every rule language should support it in order to  
be minimally RIF compliant.

This is, I guess, a reprise on that "What goes in core?" issue. But,  
as I understand it, core is a bit less than minimal, so why not err  
on the side of more?

(I'm just trying to understand the general principle here. I would  
not find it happy if, in the future, new useful datatypes for OWL  
could not be standardized at the W3C unless RIF also had them as part  
of the basic set. OTOH, I have no problem with a Profile of owl being  
designed to fit into RIF (e.g., OWL RL or OWL QL) since part of the  
point of those profiles is to hit certain sweet spots. RIF  
compatibility, arguably, is part of some sweet spots.)

Cheers,
Bijan.

Received on Monday, 26 January 2009 16:19:33 UTC