- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Sun, 03 Feb 2002 12:08:43 +0200
- To: ext Frank Manola <fmanola@mitre.org>, Sergey Melnik <melnik@db.stanford.edu>
- CC: Uche Ogbuji <uche.ogbuji@fourthought.com>, Jonathan Borden <jborden@mediaone.net>, Jeremy Carroll <jjc@hplb.hpl.hp.com>, RDF Interest <www-rdf-interest@w3.org>, RDF Comments <www-rdf-comments@w3.org>, <www-xml-schema-comments@w3.org>, Dan Connolly <connolly@w3.org>
I agree in general with Frank's approach, though I want to stress that one of the proposals does not require any of this extra machinery to capture all the required semantics -- and that we should avoid the addition of any such extra machinery if it is at all possible. It is a side-effect of the "philosophy" about datatyping taken by some proposals that such extra machinery must be created, and I would consider that a mark against the overall acceptability of such approaches -- given another approach that does the job without the addition of extra URIs, vocabulary, etc. Cheers, Patrick On 2002-02-02 21:59, "ext Frank Manola" <fmanola@mitre.org> wrote: > Sergey-- > > I think we agree. As long as we're talking about the XML schema > datatypes *as datatypes*, we ought to be able to use the XML schema > URIs. But, as I said, when RDF Core introduces new concepts, it should > label them using an RDF namespace. I should have also added, "when RDF > Core needs URIs for existing concepts that don't currently have URIs, it > should create them using an RDF namespace". That seems to be the case > with things like the value spaces of XML schema types, as you noted with > the example of http://www.w3.org/2001/XMLSchema#int.val. As you say, > perhaps we should ask for these URIs to be created (or for permission to > add them). In the meantime, it seems to me that one way to proceed, > from the "modeling" point of view, would be to define a set of RDF > predicates (in an RDF namespace) which define our view of the "standard" > relationships between XML schema datatypes and their value spaces, > lexical spaces, and any other stuff we want to be able to name in > RDF-level discussions. So we would then be able to say that something > like the following (inventing some names) holds: > > [XMLSchema#int]---RDFdatatypevalue_space--->[RDFdt#int.val] > > Here, we're preserving the meaning of the XMLSchema URL (it refers to > the datatype concept defined in XMLSchema), but adding some additional > information of our own that we happen to need (which we're always free > to do, as long as we don't claim that XMLSchema defines it). This > approach could also be a way to define the "simple and normative > mapping" that Uche Ogbuji suggested (at least, pending any action that > the XML Schema folks might take to do this for us). > > --Frank > > Sergey Melnik wrote: > >> I agree with your point. The trouble of using those genuine XML >> datatypes is that the XSD document introduces those URIs for datatypes >> as a whole, which are some kind of complex abstract objects. >> Specifically, datatypes are defined as 3-tuples, so that the URI like >> http://www.w3.org/2001/XMLSchema#int effectively denotes a 3-tuple, but >> not the value space or lexical space of the datatype. >> >> Some datatyping schemes and idioms that are currently under >> consideration require explicit identifiers for say value spaces. For >> this reason, I introduced URIs like >> http://www.w3.org/2001/XMLSchema#int.val in [1]. Using such URIs would >> be "politically correct" only if the authors of the XSD spec assign them >> explicitly the meaning we think they should carry - in fact, maybe we >> (the RDFCoreWG) could ask them to do so, or they could authorize us? I >> understand there is no written rule that prohibits us to define URIs in >> the xsd: namespace (in fact, we would just give an explicit name to >> something that's already defined in XSD). However, defining vocabulary >> in someone else's namespace feels like setting up your own Web page on >> someone else's Web server without authorization (I think DanC can argue >> better for this cause). >> >> It is still quite likely that RDF datatyping could do without new >> vocabulary. For example, we could define the class extensions (CEXT) of >> resources like http://www.w3.org/2001/XMLSchema#int as datatype mappings >> (or sets of pairs). One could claim that by doing so we just give an RDF >> interpretation for an existing concept, and not specify a new one... >> >> Sergey >> >> [1] http://www-db.stanford.edu/~melnik/rdf/datatyping/ >> >> >> Frank Manola wrote: >> >>> For the benefit of the uneducated among us, would you care to explain >>> this a bit more fully? To the extent that RDF Core truly introduces new >>> concepts, it certainly should label them using an RDF namespace. >>> However, if RDF wants to use values that have genuine XML datatypes >>> (especially if those values are going to be represented in RDF's XML >>> syntax), why should we not say "xsd:integer" rather than >>> "rdfdt:integer"? I'm not talking now about datatypes that are "sort of >>> like" XML datatypes, but are really and truly XML datatypes (as is, I >>> believe, what we're trying to do). What's the point of having URIs if >>> you have to invent new ones in order to refer to what is supposed to be >>> the same concept from a different language? Talk about a "chaos of >>> namespaces and architectures"! >>> >>> --Frank >>> >>> Sergey Melnik wrote: >>> >>>> Janathan, Uche, DanC, >>>> >>>> thank you for identifying the problem (I do remember DanC's posting >>>> related to grazing on someone else's grass ;) >>>> >>>> I'm going to replace xsd: by rdfdt: in the next revision of the >>>> document. >>>> >>> snip >>> >>>> Uche Ogbuji wrote: >>>> >>>>>> http://lists.w3.org/Archives/Public/www-archive/2002Jan/att-0131/01-RDF_D >>>>>> ata >>>>>> >>>>>>> typing.htm >>>>>>> >>>>>> I am concerned that this document element names into the XML Schema >>>>>> namespace. It seems to me that concepts that RDFCore introduces should be >>>>>> labelled by an RDF namespace. It seems to me that the XML Schema >>>>>> namespace >>>>>> should be reserved for XML elements and URIs introduced by this WG. >>>>>> >>>>> I agree with this, but I'd go farther. I think that even though RDFCore >>>>> is >>>>> not chartered to come up with a new data typing scheme, that they should >>>>> consider defining XSD data types using URIs under the control of RDFCore, >>>>> and >>>>> providing a simple and normartive mapping between these and the XSD data >>>>> types. >>>>> >>>>> I think that given the current chaos of namespaces and architectures in >>>>> the >>>>> W3C, that this is the only safe approach for consistency *within* the RDF >>>>> space. >>>>> >>>>> >>> snip >>> >>>>>> i..e. just don't call it "xsd:integer" rather "rdfdt:integer" >>>>>> >>> -- >>> Frank Manola The MITRE Corporation >>> 202 Burlington Road, MS A345 Bedford, MA 01730-1420 >>> mailto:fmanola@mitre.org voice: 781-271-8147 FAX: 781-271-8752 >>> > -- 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 05:07:36 UTC