- From: Mor Peleg <peleg@SMI.Stanford.EDU>
- Date: Tue, 26 Sep 2000 09:36:19 -0700
- To: Tom Van Eetvelde <tom.van_eetvelde@alcatel.be>, Graham Klyne <GK@dial.pipex.com>, Ray Fergerson <fergerson@SMI.Stanford.EDU>, rdf interest <www-rdf-interest@w3.org>, Gerard Maas <gerard.maas@alcatel.be>, Erik De Blieck <Erik.De_Blieck@alcatel.be>
Hi tom, Thank you very much for your answer. I am not clear about one thing: in the Bag, are the different enumerations Classes or String Literals? Thanks very much, Mor At 02:50 PM 09/26/2000 +0200, Tom Van Eetvelde wrote: >Hello RDF comunity, > >I did some more thinking as I was not pleased with my own proposal to >implement an enumeration of days. > >The core problem, I think, is the fact that the 'range' property only >accepts classes. When validating >(or maybe better: typechecking), only instances of those classes may >appear. This means that literals >can never be used as the instance must be typed, which means that it is a >resource (with a 'type' >property)! This questions the use of literals in the RDF schema model. >Literals can only be used for the >moment with properties that are not restricted in range. This makes the >example 1 of the RDF M&S a bit >akward: what you would like to use is the following: > ><s:Person rdf:ID="Mike"> > <s:maritalstatus> Single </s:maritalstatus> ></s:Person> > >and not (the way we have to formulate it now): > ><s:Person rdf:ID="Mike"> > <s:maritalstatus rdf:resource="Single"/> ></s:Person> > >I currently believe that the most elegant way to introduce enumerations is >to define a new >constraintproperty instance: 'literalRange'. This constraintproperty can >only have 'bag' instances as >value. > >Continuing the 'days' example then: make a bag with all the days that you >need. Make your property that >you want to see restricted to days and add to this property the >literalRange = days bag. When making an >RDF instance using the defined property, the type checker can search for >the filled-in value in the bag. >If it is in, validation = ok, else nok. > >Conclusion: the current defined 'RDF keywords' do not suffice to capture >all kinds of behaviour that the >user wants. The current framework doesn't really support the use of >literals when it comes to type >checking. > >Greetings, > >Tom. ******************************************** Mor Peleg, Ph.D. Stanford Medical Informatics Medical School Office Building x-208, 251 Campus Drive Stanford, CA 94305-5479 Phone: (650) 723-7711 Fax: (650) 725-7944 E-Mail: peleg@smi.stanford.edu URL: http://smi-web.stanford.edu/people/peleg/
Received on Tuesday, 26 September 2000 12:40:48 UTC