Re: [LC response] To Jeff Heflin

Hi Jeff,

If I might add a personal follow up about having various "built-in"  
operators on data values.

There are two basic ways to do this in OWL:
	1) By extension of the existing datatype support
	2) By adding some orthogonal support, e.g., in the form of rules

In the currently understood and likely to be implemented in a  
reasonable time frame forms, these have semantic consequences, i.e.:

	1) definitely can affect the subsumption hierarchy
	2) generally only works on fairly explicit values connected to named  
individuals (i.e., DL Safe rules)

Because of these differences, any near terms support of 1) will only  
be for "path free" expressions, that is, for data values that are  
directly connected to a single (possibly anonymous) individual. 2) has  
no such restriction because of its restriction to named individuals.

Then, of course, there's what operators you support.

The working group(s) made progress on all fronts in spite of some  
vehement controversy :) In particular, 2) is partially delegated to  
RIF but is also being taken up by the implementors (Pellet and HermiT,  
in particular) though several of the issues of dealing with data  
values in rules in the presences of OWL axioms were hammered out in  
the work on easy keys. For 1) we have both syntactic hooks in the  
language and a first extension (for linear (in)equations) that uses  
those hooks.

Perhaps most significantly, the core datatype support is hugely  
improved (which was surprisingly more technically and socially  
difficult than I think anyone expected). The datatype support is  
speced out to a level of detail that sets the bar appropriately high  
and covers enough types to, IMHO, stimulate users to use them  
aggressively. This will make introducing additional support  
much...well..."easier" is not quite what I mean...less speculative.

So, still more work to do, but progress.

Cheers,
Bijan.

Received on Saturday, 8 August 2009 10:49:06 UTC