Re: model theory (Alt and disjunction: just say no)

pat hayes wrote:
> 
> Attached is the final version of the MT document.

|Alternatives are transparent containers which have a special
disjunctive
|intepretation.
[...]
|>> if x is in IALT then for any z in IP, <x y> is in IEXT(z) iff
|for some n, <x(n) y> is in IEXT(z)

Let's not go there/do that.

Alt is just another class. There's nothing special about it
except that applications can treat it specially because
they know its name.

If

	_:something :relatedTo _:C.
	_:C rdf:type rdf:Alt.
	_:C rdf:_1 _:x.
	_:C rdf:_2 _:y.

then we neither license

	_:something :relatedTo _:x.

nor

	_:something :relatedTo _:y.


This is similar to the situation where folks assume that

	_:something dc:creator _:C.
	_:C a rdf:Bag.
	_:C rdf:_1 _:Kernighan.
	_:C rdf:_2 _:Ritchie.

entails

	_:something dc:creator _:Kernighan.

But it doesn't. Documentation for the dc:creator property
might say that *for that property* the above inference
is licensed. But it's not a feature of the RDF core language.

Similarly for rdf:Alt.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/