- From: Andy Seaborne <andy.seaborne@epimorphics.com>
- Date: Fri, 13 May 2011 16:17:35 +0100
- To: public-rdf-wg@w3.org
On 13/05/11 15:48, Lee Feigenbaum wrote: > On 5/13/2011 10:33 AM, Alex Hall wrote: ... >> FWIW, my preferred approach would be to: >> 1. Say that every literal has *either* a datatype *or* a language tag. >> 2. Say that the datatype of the surface form "foo" is xsd:string. > > I also prefer this approach. I don't really understand the preference > for normalizing to a plain literal with no datatype or language tag. I > know Andy talked about users wanting similarity between language tagged > literals and simple string literals, but I don't really even know what > wanting that similarity means. If, in Turtle, the input form "foo" becomes "foo"^^xsd:string and it gets written out again, as "foo", it means that it is impossible to write a simple literal (old style) at all. If "foo"^^xsd:string turns into "foo", writing old style (= existing) Turtle has the option of both forms. The test is "has datatype? no => a string of some kind". I don't mind from a technical point of view that the internal form having a datatype of xsd:string but I think users will find the fact that "foo" has a datatype and "foo"@en does not is odd. The ideal outcome here is if we had a theory of how a datatype hierarchy for language tags, with all the details of language tags like region and script work. As we don't, I suggest we avoid adding rdf:PlainLiteral into the mix but leave the space open to a detailed solution in the future. For rdf:PlainLiteral: the lexical form is confusing and it hides the language tag - practical consideration for people using RDF without reading all the specs, not a theoretical problem. Andy
Received on Friday, 13 May 2011 15:18:05 UTC