(replying to myself; pointing out my error) > of the "unique names" assumption (the idea that distinct names denote > distinct objects). > > I think we agree that it's impractical to have this assumption among > names in a global (widely distributed) namespace, so if RDF documents > are to all share one denotation map, then they must work without the > unique names assumption. ... > equality, no negation, .... Anyway, it seems to me that the terms, > rather than being constants, should all be existentially quantified > variables: that makes them match our not having the unique names > assumption. Dan Connolly pointed me to http://logic.stanford.edu/kif/Hypertext/node38.html which shows that KIF (and thus presumably FOL) does not have the unique names assumption -- it's inference systems that typically have it. So now I'm looking at http://logic.stanford.edu/kif/Hypertext/node4.html and thinking RDF layer 1 is equivalent to KIF restricted (by removing grammar productions) to this: <objconst> ::= a word denoting an object <relconst> ::= a word denoting a relation <term> ::= <objconst> | <relconst> <sentence> ::= <relsent> | <logsent> <relsent> ::= (<relconst> <term> <term>) <logsent> ::= (and <sentence> <sentence> <sentence>*) I bet this is a lot smaller than the core you have in mind for IEEE-KIF. :-) -- sandroReceived on Monday, 21 May 2001 14:53:53 UTC
This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:45:38 UTC