- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Mon, 25 Feb 2002 07:42:01 +0100
- To: ext Aaron Swartz <me@aaronsw.com>, Pat Hayes <phayes@ai.uwf.edu>, RDF Core <w3c-rdfcore-wg@w3.org>
- CC: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
On 2002-02-24 22:05, "ext Aaron Swartz" <me@aaronsw.com> wrote: > On 2002-02-24 2:54 PM, "Pat Hayes" <phayes@ai.uwf.edu> wrote: > >> ex:age rdfs:range _:x . >> xsd:integer rdfs:range _:x . >> >> If that is acceptable, then that solves the problem, since it frees >> up the datatype name itself to have the value space as its extension >> when used as a class name. > > I have no problem with the basic concept behind this, but I think that this > specific version of it isn't going to work since it's impossible to write in > RDF/XML. One way to express the above semantics, which I've proposed before with no response whatsoever from anyone, is to use a specific range property that asserts a lexical space specific constraint on property values. I.e. rdfs:lrange. It can be easily expressed in RDF/XML. It has clear, reasonably mnemonic meaning to users. It leaves rdfs:range to denote constraint to value space. It is close enough to rdfs:range that it will seem familiar while having a mnemonic distinction (l for lexical). >> Of course we cannot use rdfs:range in this case, so this would >> require us to reconsider decision (1); but the required change is >> minimal. I'm not saying yet which of the three proposals, if any, I wan't to vote for (have to sleep on it a bit) but I think rdfs:lrange would sufficiently address the objections to the expression of this particular approach. > I think Brian made it very clear that the <range> in one wasn't necessarily > rdfs:range. It could also be rdfd:range. I think this would be an adequate > solution also. That's true. Though from the perspective of the less-technical users, a somewhat mnemonic property would I think be appreciated, and the least amount of new vocabulary as possible. Using rdfs:range and rdfs:lrange provides that and covers the bases. Of course, no matter what we call the ranges or how we express them, having range constraints for both value and lexical space in the same graph will raise a conflict. This is a problem with cohabitation of the inline and bNode idioms that was pointed out since the earliest considerations of S. If we are to allow both idioms to co-exist in the same graph, and I think we must, then we will need a third "flavor" of range constraint -- one which allows members of either lexical or value space, leaving it to the graph syntax (literal or bNode) to discern the distinction. If it is a literal, it is a member of the lexcal space. If it is a bNode, it denotes a member of the value space. I think that perhaps Pat's finding of such a "union" range constraint as unworkable hinges on the fact that it treats the internal structure of the datatype as opaque, not concerning itself about whether there actually are two different spaces and what their relationship is, but just treats all members of all spaces as equal members of the *RDF* datatype class. It may very well be that this is too ambiguous for e.g. entailment proofs (though I'm not sure we have fully explored the options there) but it does meet the needs to express datatyping in both inline and bNode idioms and constrain property ranges to either flavor of idiom. I don't want to reduce the utility of RDF datatyping for WebOnt, but at the same time, don't want to see it too cumbersome and confusing for the metadata and content management communities also. And there is always the option of providing a solution now that has minimal definition in the MT and the remainder expressed in axiomatic rules and by other explicit means (even if less formal that the MT) which captures the idioms and expected meaning of them insofar as humans are concerned, and then continue to work on a more comprehensive MT treatment of datatyping in terms of the MT -- i.e. if we decide at this very moment in time to only capture pairings in the MT, per the "Dummies" proposal, that doesn't mean that we have given up on an MT account of datatyping suitable for WebOnt, only that that must come as a subsequent (possibly immediate) next step, either by us (if we still have time/energy) or by WebOnt or some other WG. Need to think on this still a little bit (as I'm sure we all do). 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 Monday, 25 February 2002 01:40:30 UTC