- From: <Patrick.Stickler@nokia.com>
- Date: Tue, 6 Nov 2001 17:22:51 +0200
- To: bwm@hplb.hpl.hp.com, phayes@ai.uwf.edu
- Cc: w3c-rdfcore-wg@w3.org
> There seem to be three fundamentally different approaches in > play. They all > have in common that a literal value, e.g. an integer, is > denoted by a node in > the graph. They differ on whether an arc from that node, > labelled with > rdf:type, takes a value which denotes a value space or a > datatype (Pat's use of > the term datatype i.e. as defined in XSD a tuple defining > lexical space, value > space and facets). > > All I was originally trying to say is that programmers, and > some of the time I > am one, are used to the type of something denoting the value > space alone. > > Brian It is true that programmers are used to dealing primarily only with the value spaces, but that is because the lexical forms of values are not preserved, but only a means to an end, that end being a canonical internal representation of the value. Since RDF preserves the lexical form, we cannot enjoy such a convenience. At least not at the RDF level. Perhaps in the future, we will have logical levels that will provide such transparency of lexical form, just as we envision having transparency of ontological terms with equivalent semantics, but that is not going to happen at the RDF (i.e. graph) level itself. So the solutions that we come up with must at least preserve the needed information for interpreting the lexical forms of literals as well as organize that information in a consistent manner for the sake of implementation. Right? Patrick
Received on Tuesday, 6 November 2001 10:23:12 UTC