- From: Brian McBride <bwm@hplb.hpl.hp.com>
- Date: Tue, 24 Jul 2001 12:09:10 +0100
- To: pat hayes <phayes@ai.uwf.edu>
- CC: w3c-rdfcore-wg@w3.org
pat hayes wrote: > OK, though you really shouldnt have used the same '-:q' as in the > first example. I did that exactly to expose the issue of the scope of anon resource names. > BUt now, how is any matching done between these? > Rewrite the second example using a different anon-name: > > ><#advert456> <foo:role> "seller" . > ><#advert123> <foo:description> _:service . > >_:service <foo:produce> <foo:roses> . > >_:service <foo:quantity> _:qq. > >_:qq <foo:units> <foo:kg> . > >_:qq <foo:maxValue> "500" . > > and put it all together and we have a total of 12 triples which refer > to two anonymous things called _:q and _:qq which have no obvious > connection to one another. No variables, nothing to 'match', no > obvious consequences. Certainly it doesnt follow that _:q equals > _:qq; they might be respectively 256 and 483, say. It was never my intentiont tht _:q and _:qq denote the same resource. One denotes a buyer service. The other denotes a seller service. However, a broker might usefully note some things in common between these two descriptions. A broker might 'know' that there is profit to be made introducing two services, one a buyer and the other a seller of the same product where the quantity constraints are compatible. > Agreed, though that isnt the key issue here. What matters is what > kind of process is allowed to 'bind' a value to a variable, and the > (sad but true) fact that whatever is doing that had better not claim > to be doing anything that is valid in RDF1.0. On what grounds? > > > > >[...] > > > > I'm not sure that standard FOL captures this. FOL is built > > > > around a conceptual model where there can be many interpretations > > > > for statements in the FOL. But that is not the situation we > > > > are in here. We have one interpretation - its a mapping to > > > > the world out there. > > WHICH mapping? How can I possibly know, when reading your RDF, which > mapping you have in mind? Are you suggesting that there is more than one mapping from a URI to a resource? From a URI to a property? I hope not. [...] > > The key issue is whether something is allowed to bind variables to > new values during processing. If they are, and if this is considered > valid, then those variables that get bound must be either universal > variables in an assertion, or existential variables in a query. > Either way that takes us outside the RDF 1.0 M&S, seems to me. > > > > >What I'm getting at that traditional logic is about designing formal > >systems that are true under any interpretation. That's not what we > >are dealing with here. > > No. Logic is about characterising how truth is preserved. The idea > isnt that formal systems are true (actually that doesnt mean > anything), but that the formal systems *preserve* truth: it the > antecedents are true (in I) then the consequents must be true (in I). > That 'if..then' is what has to be true for any I if the reasoning is > valid. The point being that if I have no idea what your intended > interpretation is (which in general I don't, in fact, other than > knowing that it makes your assertions true), then that doesnt matter; > I can still draw valid conclusions from your sentences since > *whatever* your intended interpretation was, my conclusions will be > true in it. Quite right I wrote with my usual fuzzy lack of precision. However, the point I was trying to make remains valid. RDF is not about a truth preserving logic system which preserves truth in any interpretation. Its about expressing facts with one interpretation. > The key issue is who is doing the binding. If an RDF engine can bind > values to variables at run time (ie at inference time), then it's > going beyond the M&S. Does M&S preclude this? Brian
Received on Tuesday, 24 July 2001 07:11:45 UTC