- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Sun, 03 Feb 2002 11:40:41 +0200
- To: "ext jos.deroo.jd@belgium.agfa.com" <jos.deroo.jd@belgium.agfa.com>, RDF Core <w3c-rdfcore-wg@w3.org>
On 2002-02-03 1:03, "ext jos.deroo.jd@belgium.agfa.com" <jos.deroo.jd@belgium.agfa.com> wrote: > ??? why not take just the 2 > > dt-local = S-A + TDL-local > > :Jenny :age [ xsd:int.map "30"; rdf:dtype xsd:int.val ] . > > dt-global = S-B > > :Jenny :age "30" . > :age rdfs:range xsd:int.lex . > > -- > Jos De Roo > > Firstly, the above is not TDL-local. TDL-local uses rdf:value to bind the literal (lexical form) to the bnode and the value of rdf:type or rdf:dtype is the URI of the datatype, not the value space. Secondly, the above approach doesn't really do anything more than TDL-local and TDL-global except require that we have 4 URIs for each datatype rather than one and preclude use of both idioms together (see next) Thirdly, having the range of a literal node as *.lex and the range of a bnode *.val is precisely the problem that S has with cohabitation of its local and global idioms. One cannot then define a range for the local idiom intended to express a constraint without conflict of interpretation (is the bnode *.lex or *.val?) There is already alot of negative opinion about how the RDF datatyping proposals are referencing XML Schema datatype URIs. I suspect that any solution that uses anything other than the pre- defined (or user defined) single URIs for each datatype will not be acceptable. The TDL approach of having the datatype URI denote the entire datatype, not just one of its components, seems to sync the closest with the desires of the XML Schema and "don't graze on other's grass" folks. I think one key difference between the TDL and S "philosophies" is that S wants/needs to use a unique URI for each component of a datatype in order to make the significance of those components explicit in the representation, whereas TDL uses the single URI of a datatype to define a context within which the MT provides a consistent interpretation for the lexical form (literal). I concede (as I always have) that having an MT that works and meets folks expectations/needs is imperative -- but much of our discussion seems to be focusing solely on the state of the MT and does not sufficiently consider usability, efficiency of expresssion, and in short, what the ramifications will be for "common users". Multiple URIs, synonymous idiom-specific vocabularies, etc. etc. may make the MT easier to write, but it makes life in general much harder for the user, and after all, at the end of the day, if RDF datatyping is percieved to be too complicated, regardless of how beautiful and correct the MT is, folks won't use it. Eh? The S approach makes things easier for writing the MT but harder for the users. Who should we be trying most to make things easiest for? I think the users. Please let us ask ourselves, not simply does it work, but does it make life easier for the majority of folks who want to use RDF for common metadata and knowledge management tasks. And the majority of folks are not necessarily those building complex KR or expert systems to explore theorem proving, etc. but folks who simply want to describe things and find things based on those descriptions. RDF is already percieved as difficult and confusing enough -- we don't need to add to that perception by having a cumbersome datatyping solution (if we can by any reasonable means avoid it). I think we should perhaps add an item to the desiderada, that the solution be as simple and easy to use as possible, even at the expense of a more complicated MT. Cheers, Patrick -- Patrick Stickler Phone: +358 50 483 9453 Senior Research Scientist Fax: +358 7180 35409 Nokia Research Center Email: patrick.stickler@nokia.com
Received on Sunday, 3 February 2002 04:39:39 UTC