ISSUE-93: datatype IRI mapping

Over the coming weeks as we wrap-up the RIF WG we need to identify which issues 
we are going to address and which will be tabled (postponed).  This is the same 
process we went through for BLD last call.

Here is my summary of the status of ISSUE-93 (Should datatype IRIs map to the 
datatypes themselves), which we discussed at the last telecon. The issue is that 
the current treatment of datatype IRIs, which became clear as we added the 
pred:isLiteralOfType (& al) predicates, sort of treats them as primitive literal 
IRI strings, ignoring the general treatment in RIF of rif:iri (that they can 
denote anything).  Thus pred:isLiteralOfType expects its argument to be 
something like xsd:string^^rif:iri, but that IRI, as a rif:IRI, could in 
principle be made to denote anything, e.g. <xsd:string> = <xsd:int> or something 
else.

Jos suggested we change the definition of IRI constants so that certain IRIs, in 
particular the ones that we take to denote xsd datatypes, denote the things we 
mean.  This would require changing the semantics in BLD.

Michael did not like changing the general definition of IRI for a few 
exceptions, proposing to change pred:isLiteralOfType so that it takes a argument 
of type anyuri, which is a literal. This would require changing the definition 
of the four "meta" predicates, which haven't appeared in a formal WD yet.

Axel suggested, though not that strongly, that the current approach is sort of a 
compromise between these two, and yes if someone said <xsd:string> = <xsd:int> 
then too bad, you should know not to do that.

</chair>
Technically I prefer Jos' solution.  Administratively I prefer Michael's, as it 
doesn't require changing BLD.
<chair>

-- 
Dr. Christopher A. Welty                    IBM Watson Research Center
+1.914.784.7055                             19 Skyline Dr.
cawelty@gmail.com                           Hawthorne, NY 10532
http://www.research.ibm.com/people/w/welty

Received on Friday, 27 February 2009 16:58:20 UTC