- From: Michael Kifer <kifer@cs.sunysb.edu>
- Date: Fri, 05 May 2006 15:33:23 -0400
- To: Francois Bry <bry@ifi.lmu.de>
- Cc: public-rif-wg@w3.org
> > Michael Kifer wrote: > >> Do you really mean this, i.e., that there is essentially only one domain of > >> discourse? Are there no possibilities of non-trivial identity, for example > >> between f(a) and f(b)? > >> > > > > With equality the domain of interpretation is a set of equivalence classes > > over the Herbrand universe. Herbrand universe != Herbrand domain in case > > there is equality. This is all standard stuff in Logic Programming. > > http://citeseer.ist.psu.edu/context/6262/0 > > Also see the classical Chang & Lee's book: > > > > @book{ chang-lee, > > author = "C.L. Chang and R.C.T. Lee", > > title = "Symbolic Logic and Mechanical Theorem Proving", > > publisher = "Academic Press", > > year = 1973 > > } > > > > > > To the beat of my understanding, things can be a little bit more > differentiated. > > There are two approaches to equality in logic programmi9ng and rule > languages. > > - the simple and limited treatment of equality as "syntactical > equality". In this case, f(a) != f(b) and the (in)equality between > expressions containing variables is only allowed if these variables are > bound when this expression is evaluated. This like "no equality". In standard Prolog the equivalence classes are each singleton. > > - the more complicated one: what Michael has mentioned above. This > approach is a full treatment of equality at the cost of a significantly > more complicated reasoning procedure (cf. "paramodulatrion"). Peter was interested in non-trivial equality, I am sure, as in DLs. --michael > Many rule languages, especially Prolog and most PR and RR languages, > have only equality of the first kind. > > Francois > > >
Received on Friday, 5 May 2006 19:33:43 UTC