Re: RIF-BLD Classification

> 
> 
> Michael Kifer wrote:
> >> The principle argument for rif:subclassOf is that it is more amenable 
> >> to interchange since it is minimal, as well as a natural part of usual 
> >> frame style syntax.
> > 
> > rif:subclassOf (aka ##) also has standard semantics as in all
> > object-oriented languages.
> 
> Yes, although note the semantics of the proposed rif:type is not the 
> same as the semantics of some popular OO languages (e.g. Java), as 
> Gary & Dave pointed out on the telecon.  You would need closed-world 
> negation and some restrictions on rule heads for that.  So I left out 
> that rationale here as I figured rif:type and rif:subclassOf were a 
> package deal.

Not sure what you mean by "package deal". It is not true that typing needs
CWA. It depends what you want to do with typing. RDF and OWL do not use
CWA, but have type specifications. What this means is that the
reasoning that can be done there is type inference, not type checking.
OO languages use type checking. For the latter you need to view type
specifications as constraints, not as inference rules. This is orthogonal
to CWA I would say, as there have been formulations that introduced
constraints in DL framework (the latest WWW paper by Motik et al, for
example).

Regarding the restrictions on the rule heads, this has little to do with
type checking per se. You need such restrictions if you want to do *static*
type checking. But the "static" aspect is not semantic, but algorithmic.


	--michael

Received on Friday, 3 August 2007 18:13:25 UTC