- From: Michael Schneider <schneid@fzi.de>
- Date: Sat, 1 Dec 2007 21:49:13 +0100
- To: "Pat Hayes" <phayes@ihmc.us>
- Cc: "Jeremy Carroll" <jjc@hpl.hp.com>, "Owl Dev" <public-owl-dev@w3.org>, "Ian Horrocks" <horrocks@cs.man.ac.uk>, <bmotik@cs.man.ac.uk>
Hi Pat! Pat Hayes wrote on Tuesday, November 27: >I wrote: > >>What comes to my mind is what's actually missing are "RDF >comments", i.e. >>comments which belong to an RDF graph, but which are not >interpreted (all >>triples within an RDF graph are always interpreted according to RDF >>semantics). > >Webont/RDFWG discussed this idea. They were >referred to as "dark triples", ie RDF triples >which are not required to conform to the RDF >semantics. (They have several other potential >uses in addition to commenting: at one time it >was thought that they were required to allow OWL >to be encoded into RDF.) However, they raise many >other issues, most importantly being how an RDF >processor is supposed to be able to distinguish >them from ordinary triples. To me, the most >cogent argument against this idea is that it >doesn't seem to actually be necessary. Suppose it >were somehow arranged, and then someone >accidentally thought that a dark triple were a >real RDF triple. What harm would be done? Since >RDF is so weak, I suggest that the answer is: no >harm at all. The only thing that would follow >from (be RDF-entailed by) such a dark/comment >triple would be that the comment exists (i.e. the >same triple with the comment replaced by a >bnode); and since the comment is presumably a >string, this seems a harmless entailment in any >conceivable circumstance. [snip] >>Ah, and please don't ask me about any technical details of such an RDF >>comment feature! > >Well, but it might be worth thinking about it. It >is rather a tar-pit. Is it *really* this >important to avoid semantics? A commenting triple >asserts... something (and that a character string >exists, but we knew that already.) What, exactly, >is not specified. Finally I found some time to think about this, and I agree with you. At least as long as we create only string comments of the form: <someEntityOrAxiom> rdfs:comment "just a string" . This is really a very weak "assertion". A string literal (either plain or typed as 'xsd:string') simply denotes the string value itself. So there is probably not much information one can gain from looking at such a string comment. I can, however, see some subtle source of danger, which should at least be mentioned once: In OWL-Full, people are allowed to assert a "<=1" cardinality restriction on the 'rdfs:comment' property. Say, I have two different URIrefs denoting individuals, and two different string comments, as in <individual1> rdfs:comment "first comment" . <individual2> rdfs:comment "second comment" . then the "<=1" cardinality restriction will allow me to conclude <individual1> owl:differentFrom <individual2> . I normally would not expect "semantically weak" comments to be responsible for such a strong entailment. But, of course, I would call putting such a cardinality restriction on a property, which is intended for annotation purposes only, to be clear abuse. >So, is this really so dangerous >that we need to revise the world to avoid it? No, probably not. If I had to choose between working on a big revision of RDF on the one side, and writing down (maybe in a "best practices" document) a set of reasonable pragmatics like "Do not put cardinality restrictions on rdfs:comment!", I would definitely decide myself for the second option. I can see that people want to have annotations for both entities and axioms, and it is not hard for me to believe that such annotations are really useful. And an RDF-mapping, which translates such annotations to string comments, does not seem to be harmful in /practice/. The next two questions will then be, on the one hand the question about more general comments having arbitrary objects, and on the other hand the question about being able to create sub properties of annotation properties. Jeremy's idea to use the universal property sounds kind of fascinating to me, but I feal a little uneasy with this approach, without being able to explain... Well, fact is that I had no time yet to think more deeply about this, so I will stop here for the moment. :) Cheers, Michael -- Dipl.-Inform. Michael Schneider FZI Forschungszentrum Informatik Karlsruhe Abtl. Information Process Engineering (IPE) Tel : +49-721-9654-726 Fax : +49-721-9654-727 Email: Michael.Schneider@fzi.de Web : http://www.fzi.de/ipe/eng/mitarbeiter.php?id=555 FZI Forschungszentrum Informatik an der Universität Karlsruhe Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe Tel.: +49-721-9654-0, Fax: +49-721-9654-959 Stiftung des bürgerlichen Rechts Az: 14-0563.1 Regierungspräsidium Karlsruhe Vorstand: Rüdiger Dillmann, Michael Flor, Jivka Ovtcharova, Rudi Studer Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
Received on Saturday, 1 December 2007 20:49:40 UTC