- From: Sandro Hawke <sandro@w3.org>
- Date: Fri, 18 Nov 2011 09:01:07 -0500
- To: Andy Seaborne <andy.seaborne@epimorphics.com>
- Cc: public-rdf-wg@w3.org
On Fri, 2011-11-18 at 13:29 +0000, Andy Seaborne wrote: > > On 17/11/11 21:27, Sandro Hawke wrote: > > On Thu, 2011-11-17 at 20:32 +0000, Richard Cyganiak wrote: > >> On 17 Nov 2011, at 19:33, Sandro Hawke wrote: > >>> I think it should actually be this group which does the revision, > >>> though. > >> > >> Why? > > > > Because RDF should be the common data model. > > > > rdf:PlainLiteral was invented because a group of people trying to build > > on top of RDF found the data model just too broken to work with. (They > > needed something simple and consistent enough on which to build > > additional logics.) So, together with everyone who was willing to help, > > we came up with something that isn't pretty but that does work. > > > >> The rdf:PlainLiteral spec defines a datatype that's defined as being usable only *outside* of RDF graphs. > > > > That's not a design goal, it's just an aspect of how we had to define it > > to not break anything existing. Given RDF 1.1 is willing to make some > > changes to existing things, it wouldn't necessarily have to be done that > > way any more. > > s/break/change/ I don't quite see the difference in this case. > rdf:PlainLiteral appear very late in the OWL/RIF WG cycle. It was called rif:text for a long time. It wasn't until the OWL 2 WG decided it wanted something similar that we started trying to come up with a shareable name for it. > > > >> It defines facets for that datatype. It defines XPath functions. None of these things are directly useful for RDF. They all make sense for, and are motivated by, RIF and OWL. > > > > If you're going to actually define a useful XML datatype, it makes sense > > to define facets and XPath functions for it. The fact that OWL 2 can > > use the facets and RIF can use the XPath functions helped motivate it, > > but it seemed like one would want these things anyway. I mean, if you > > have a language tagged string, don't you want to be able to constrain > > and/or act upon the language tag? So we defined a way to do that which > > happens to fit neatly into the existing XML datatype mechanisms. They > > are in no way just for RIF or OWL -- they are for anyone who wants to > > use strings with language tags in RDF, using machinery from the XML > > world. > > > > Maybe XML is dead, so this doesn't matter any more? > > XML treats language tag as an orthogonal to datatype > through xml:lang and through schema datatyping. > > In that sense, RDF-2004 reflected XML quite accurately. Perhaps, but having another orthogonal component has significant cost. My point was that *if* you're going to make a datatype that embeds the language tag into the value, *then* in the XML world it makes sense to provide facets and functions to get at the text and the language tag. > rdf:PlainLiteral uses a 1-D lexical space for a 2-D value space. The > lang tag does not have the same status as the datatype, > > Andy So, I'm not quite sure what you're arguing here. I think the issue at hand is what role the RDF WG should take in updating the rdf:PlainLiteral spec. I think we could reasonably do anything on the spectrum from (1) mention to the other groups they might want to do something to (10) update the spec ourselves. I lean closer to (10) but not strongly enough to take on the work myself. Mostly I'm happy to have it done however it's most likely to get done. -- Sandro
Received on Friday, 18 November 2011 14:01:21 UTC