- From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- Date: Wed, 30 Jan 2002 11:39:27 -0000
- To: <jos.deroo.jd@belgium.agfa.com>
- Cc: <w3c-rdfcore-wg@w3.org>
> I don't follow Jeremy [ snip ] > > nothing can be deduced w.r.t. values > i.e. > dt1(v1) = dt2(v2) > does *not* mean that v1 = v2 Within any logical datatyping framework of course it does not. My point is that S-B is explicitly not a datatyping framework. It is a framework for string manipulation, just like RDF M&S. S-B does not allow the distinction (within its theoretical model) between dt1(v1) and v1. RDF M&S is an untyped framework. S-B reads M&S as meaning every literal is a string. S-B maintains backward compatibility with this view. Thus S-B within its own terms, only talks about strings. Within S-B there is only one possibly denotation of a string, itself. Within those terms errors will not happen. But .... applications want integers as well as strings. Inevitably they will introduce more than one mapping function, I take your dt1 to map strings to strings, and dt2 to map strings to integers. S-B may say these are the application programmer's doing, and wash its hands of the matter. But the range constraints in S-B appear to license such coversions between strings and integers above the model theory. The non-expert application writer may then make type mistakes. My view is that a datatyping system should assist the application writer in not making type mistakes. Jeremy
Received on Wednesday, 30 January 2002 06:39:02 UTC