- From: Graham Klyne <gk@ninebynine.org>
- Date: Thu, 17 Jul 2003 13:02:34 +0100
- To: Brian McBride <bwm@hplb.hpl.hp.com>
- Cc: rdf core <w3c-rdfcore-wg@w3.org>, i18n <w3c-i18n-ig@w3.org>, Martin Duerst <duerst@w3.org>
At 11:58 17/07/03 +0100, Brian McBride wrote: >On Tue, 2003-07-15 at 16:48, Graham Klyne wrote: > >[...] > > > > > My impression is that no showstopper has been identified, but the current > > approach will be quite painful for some. > >Have we identified whom? Since you ask, application developers, I guess (more than tool developers). If I were writing an RDF application that dealt extensively with properties containing human readable text (e.g. properties like rdfs:comment or foaf:name, which I think might include applications like your meeting assistant), which might also reasonably contain some kind of additional markup (HTML tags, Ruby?), then I think that I would find having two different ways of handling language tags would be a real pain; it would create alternative code-paths where there should really be just one. Having to add things like the suggested <span> elements to convey language information would, I think greatly complicate some of those paths to the extent that some developers might be tempted to skip handling those forms of data that might be needed for effective I18N. ... I started in this debate feeling quite neutral, but the more I think about it the more I find that I dislike the special treatment of XML in RDF. All of the actual RDF applications I've contemplated use literals in one of two ways: (a) as human-readable text, and (b) as simple data values with their own formatting rules and other properties. In the past, I never really considered there to be an important role for XML *within* RDF literals, and have been happy enough to let others drive that aspect of the design in ways they thought useful. But the proposition that XML in RDF literals is provided to extend the range of human-readable text is one that I find compelling, particularly in the face of I18N, and one that doesn't really sit comfortably with the current design (for reasons like those noted above, and in Pat's earlier message [1]). I think our datatyping design does gives us a very clear and direct way to avoid this problem of "how shall we treat XML - as text with markup or as a formal data representation language?", as I noted previously [2]: [[ Finally, I observe that dropping XML literals from the RDF specification does not preclude the later introduction of XML literals as currently defined -- they are simply another datatype. The difference would be that said datatype is not automatically signalled by the presence of parseType="Literal". ]] The treatment of any XML as anything other than text-with-possible-markup can be indicated, as with any other form of literal text, by the application of a datatype. ... I also find myself revisiting the question: what is the purpose of language tagging in RDF? I think we agreed some time ago (can't find record) that the purpose of a language tag was not for presentation, but for recording additional information about the content. To the extent that human readable text is opaque to automated systems (cf. social meaning debate), I'm wondering why it's so important at all. Ho hum. #g -- [1] http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2003Jul/0067.html [2] http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2003Jul/0155.html ------------------- Graham Klyne <GK@NineByNine.org> PGP: 0FAA 69FF C083 000B A2E9 A131 01B9 1C7A DBCA CB5E
Received on Thursday, 17 July 2003 09:30:33 UTC