- From: Graham Klyne <GK@Dial.pipex.com>
- Date: Sat, 09 Dec 2000 18:30:55 +0000
- To: "McBride, Brian" <bwm@hplb.hpl.hp.com>
- Cc: www-rdf-interest@w3c.org
At 05:11 PM 12/8/00 +0000, McBride, Brian wrote: >Pierre-Antoine, > > > about 3. I already submitted the idea below on the list, > > but I will go further : the problem raised by Jonas is not a > > problem ! If we do consider that the reified statement is > > really a statement rather than a stating, then the date > > should not be a property of st1, but rather of st2 and st4 ! > > > > st1: [Ora, creator, page] > > st2: [st1, saidBy, Ralph] > > st3: [st2, at, 01/12/99] > > st4: [st1, saidBy, Pierre-Antoine] > > st5: [st4, at, 01/12/00] > > > > The statements st2 and st4 are actually statings, not because > > *every* reified statement be a stating, but because of the > > particular meaning of their predicate "saidBy". > >Interesting approach. I'm uncomfortable with st2 - you say its >a model of a stating. But its also, by definition a model of a >statement. I've been thinking that statements and statings are >disjoint, so how can this be? Are they disjoint? Can the same >thing model both a statement and a stating? I see no reason why not (though the final answer may depend on the form of explanation (stating?) used -- so your call for formal language seems timely). In the above, I think the line starting with 'st2:' is a description of a statement known as 'st2', which itself is an assertion of a stating of the statement known as 'st1'. >We need some more formal language here - its too confusing >otherwise. > >What's key about the concept of a >stating? Basically, its a multi-valued relation - e.g. > >(stating, stmt, location, by, time, weather-conditions,...) > >or in predicate terms: > > stating(stmt, location, by, time, weather-conditions, ...) > >In RDF we only have binary predicates, so this becomes > > type(x, stating) >& stmt(x, ...) >& location(x, ...) >& by(x, ...) >& time(x, ...) >& weather-conditions(x, ...) >& ... > >Can we deduce this from st1, st2, ...? Are there a set of axioms >that would enable that deduction. And is it worth the computation >over representing things more directly? To answer the first question, I think the answer should be "yes". As to the second, I would say that implementers are free to choose an internal representation that suits their computational needs. For the time being, RDF is our assumed form of interchange and I don't feel it's productive to consider alternatives at this time if RDF is adequate. But if RDF is broken... -- I like the approach you take above above... it puts me in mind of the idea the a reification quad is the _minimum_ needed for a structure to qualify as a description of a statement, and arbitrary additional details may be added to create a richer description of the statement. I think this tends to loosen the identity between reification quad and statement, while retaining the essential elements. #g ------------ Graham Klyne (GK@ACM.ORG)
Received on Sunday, 10 December 2000 10:20:46 UTC