Re: MV Keys

In article <WjyPf.43786$Jd.38191_at_newssvr25.news.prodigy.net>, brian_at_selzer-software.com says...
> > I disagree. A value in itself is not a fact. In the context of the RM,
> > only tuples in relations are facts. The cartesian point (2,3) is not a
> > fact; the numbers 2 and 3 don't become more associated with each other
> > just because I refer to this point. Only when it is put into a relation
> > can it become (part of) a fact, e.g. "A FooBar is located at (2,3)".
> >
> You're right. A value in itself is not a fact. Perhaps I should be a
> little more precise. A composite value is a sentence which, once stated in
> a database, becomes a fact. "x is 2" is a sentence; "y is 3" is a sentence.
> "x is 2 and y is 3" is also a sentence. The juxtaposition of 2 and 3 in the
> ordered pair representing a cartesian point is a sentence. Furthermore, in
> the RM, attribute values are facts, not just tuples, because all stated
> values in the RM are named.

I disagree that names have much to do with it.

> Thus a tuple in a relation R{A, B, C} is the
> set of propositions {A has value 3, B has value 7, C has value 2}.

No. The relation R is a predicate with three variables. A tuple is a variable binding that makes the predicate evaluate to true.

Of course, this is by some definition (a common one, I believe---but perhaps not universally accepted?), so you could claim that your definition is better---but I don't think it is. For one thing, it runs into trouble the moment you introduce more tuples: If one tuple means (among other things) that "A has value 3" is a fact, and another claims that "A has value 5" is a fact, don't they contradict each other?

> The
> tuple itself also has a truth value, "A has value 3 and B has value 7 and C
> has value 2." which is a proposition in conjunctive normal form.

No, a tuple by itself is not a fact either. Only within the context of a relation does it make sense to talk about facts.

> The problem is that the widgit list domain is a set of lists of widgits,
> which can be constructed using the combinatorial rules. Some of the
> constructed lists may not violate the constraints, so the meanings overlap.
> This is not true of the use of integers in multiple places, because integers
> describe one aspect of a domain, they're not the domain.

I'm not able to make much sense of that, and I'm still not completely clear on what combinatorial rules are. Can you make a concrete example of widgit lists that shows the problems you imagine?

-- 
Jon