Re: log:conclusion, log:notIncludes, and infinite formulas

On Thu, 2005-08-25 at 11:38 -0400, Yosi Scharf wrote:
> Dan Connolly wrote:
> 
> >We all know this sends cwm into the weeds...
> >
> >@keywords is, of, a.
> >@prefix : <#>.
> >@prefix log: <http://www.w3.org/2000/10/swap/log#>.
> >
> >fred a Person.
> >{ ?WHO a Person } => { ?WHO father [ a Person ] }.
> >
> >Now suppose that's in <infaux.n3> and consider:
> >
> >@keywords is, of, a.
> >@prefix : <#>.
> >@prefix log: <http://www.w3.org/2000/10/swap/log#>.
> >
> >{ <infaux.n3>.log:semantics log:conclusion [
> >	 log:notIncludes { sky color blue } ] }
> >  => { thisTest a Pass }.
> >
> >
> >That seems like it should pass, at least in a backward-chaining
> >reasoner.
> >
> >If that's the case, then it's wrong to say
> >that the domain of log:notIncludes is finite formulas,
> >as in...
> >
> >[[
> > Because a formula is a finite size,
> > [...]
> >]]
> > -- section "Implementing defaults and log:notIncludes" 
> >   of part "Reaching out onto the Web"
> >   of the Semantic Web Tutorial Using N3
> >   http://www.w3.org/2000/10/swap/doc/Reach#Implementi
> >
> >
> >  
> >
> I don't understand what you want.

I want this aspect of the N3 semantics decided and documented.

>  Cwm is not a backwards chainer, and
> would have no way of doing that correctly.

Right; cwm is not a complete reasoner. It might be nice
to document the extent of its completeness/incompleteness,
but that's for a separate bug.

>  Do you want to commit cwm to
> being able to handle that kind of rule? Do you want to say that an
> implementation MAY return a value in the case even with an infinite formula?

More like the latter. We might need a new class of test: one that's
correct but isn't expected to be passed by cwm.

> Yosi
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Thursday, 25 August 2005 17:08:49 UTC