- From: Pat Hayes <phayes@ihmc.us>
- Date: Wed, 6 Apr 2011 15:17:03 -0500
- To: Antoine Zimmermann <antoine.zimmermann@insa-lyon.fr>
- Cc: public-rdf-wg@w3.org
On Apr 6, 2011, at 12:34 PM, Antoine Zimmermann wrote: > Le 06/04/2011 18:22, Pat Hayes a écrit : >> >> On Apr 6, 2011, at 10:19 AM, fensel wrote: >> >>> At 13:39 06.04.2011, Pierre-Antoine Champin wrote: >>>> On 04/06/2011 05:14 AM, Pat Hayes wrote: >>>>> We might want to think about incorporating some version of sameAs >>>>> into RDFS, as this seems to be fundamental to linked data and also >>>>> widely misused. Having the real meaning of equality exposed in the >>>>> RDF standard itself might be doing the world a favor. (?) >>>> >>>> +1e99 >>> >>> I think this may be a very bad idea. You would force all languages >>> layering on top of RDF to include equality. >> >> It depends what you mean by 'include'. A language based on RDF can always declare that it will use some other term and refuse to accept the RDF one, just as OWL uses owl:Thing rather than rdf:Resource. BUt it would be more useful and more in the spirit of interoperability to use the same term and just acknowledge that it is not using all the intended meaning of that term. > > I'd say that the comparison with OWL is not particularly appropriate. OWL Full is an extension of RDFS and as such, preserves the meaning of all the RDF/RDFS vocabularies. rdfs:Resource means the same in OWL Full. owl:Thing does not replace it. > > OWL DL is not an extension of RDFS and it does not try to replace rdfs:Resource by owl:Thing. rdfs:Resource simply has no meaning in OWL DL while owl:Thing has a much more specific meaning than "the class of all resources". OWL DL is very different thing from RDF which just happens to be serialisable in RDF. Yes, I know. But intuitively, rdfs:Resource means 'anything' and in OWL DL, owl:Thing also means 'anything'. They both refer to 'the universe'. The fact that RDFS and OWL-DL have different ideas of what is allowed in the universe is unfortunate, but in practice is ignored by virtually everyone, I am sure. It is just a strange idiosyncracy that these two notations insist upon using different names for the universe, is how people see it. (BTW, I have tried to explain this 'two views of the universe' idea to many people. It is not easy to understand. The analogous fact about axiomatic set theory was called a 'paradox' until the 1950s.) >>> There are reasons >>> to prevent equality because it turns unification from a syntactical >>> operation into reasoning. >> >> Only if you claim to be logically complete. A reasoner can always just ignore the equalities, or use them in a limited but useful way. Reasoners are not *obliged* to squeeze every last drop of meaning from all the RDF they encounter. Still, the RDF means what it means :-) > > I don't think this is a good way of doing things in a standardisation working group. People can always disregard standards or parts of it, of course. But going further in this direction, why not adding the whole OWL 2 vocabulary to RDF since reasoners are not obliged to implement it? Because there is no case to do so. But ask yourself: suppose that the SW 'package' had been presented to the world as a single vocabulary with one specification defining it all, and none of these variations and alternatives, but with an explicit instruction that it is fine to deploy partial implementations which must parse everything but are not obliged to have complete reasoners. This would have been MUCH easier to understand, and use, than the current incomprehensible muddle. > The WG should not define things that have a risk of being disregarded by implementers. What does this mean? Obviously, anything that can input and parse RDF triples will be able to input and parse triples containing rdf:sameAs. Not all of them will be able to draw all valid conclusions from this RDF. But very few RDFS engines can draw *all* valid conclusions from RDFS Input, even now; and the world is full of RDF that contains parts of OWL and parts of RDFS (and parts of DC and FOAF, for that matter) without anything being able to fully extract all the intended meaning from them (ie without being inferentially complete for RDFS + OWL + DC + FOAF) > If one implements a system that deals with RDF, it's ok if it includes bits from RDFS but not all. Similarly, if an implementation is doing RDFS reasoning, it's ok if it include bits of OWL reasoning. But if one wants to define an extension of RDFS, it's not ok that a portion of it is disregarded. Why not? A reasoner that only checks subClass and type information and nothing else might be useful for many purposes. It is certainly not illegal according to the specs. Nothing in the RDF spec suite requires reasoners to be complete. Pat > So, I'd say that adding owl:sameAs to RDFS is a pretty strong extension that needs to be pondered carefully and I find Dieter's concern quite reasonable. > > > > AZ >> >> Pat >> >>> >>> Dieter >>> >> >> ------------------------------------------------------------ >> IHMC (850)434 8903 or (650)494 3973 >> 40 South Alcaniz St. (850)202 4416 office >> Pensacola (850)202 4440 fax >> FL 32502 (850)291 0667 mobile >> phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes >> >> >> >> >> >> > > > -- > Antoine Zimmermann > Researcher at: > Laboratoire d'InfoRmatique en Image et Systèmes d'information > Database Group > 7 Avenue Jean Capelle > 69621 Villeurbanne Cedex > France > Tel: +33(0)4 72 43 61 74 - Fax: +33(0)4 72 43 87 13 > Lecturer at: > Institut National des Sciences Appliquées de Lyon > 20 Avenue Albert Einstein > 69621 Villeurbanne Cedex > France > antoine.zimmermann@insa-lyon.fr > http://zimmer.aprilfoolsreview.com/ > > ------------------------------------------------------------ IHMC (850)434 8903 or (650)494 3973 40 South Alcaniz St. (850)202 4416 office Pensacola (850)202 4440 fax FL 32502 (850)291 0667 mobile phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
Received on Wednesday, 6 April 2011 20:17:41 UTC