Re: The MySQL/PHP pair

"Paul" <paul_at_test.com> wrote in message news:418be676$0$561$ed2619ec_at_ptn-nntp-reader03.plus.net...
> Dawn M. Wolthuis wrote:
> >> But the predicate for this would be something like:
> >>
> >> Person P has email addresses of E1, E2, E3, ...
> >
> > or
> >
> > Person P has the following list of e-mail addresses: L
>
> You could say this. But then L is a "variable" (in the predicate logic
> sense) and not a predicate. So when we translate over to relations, L is
> atomic in the sense that the relational system cannot look inside it -
> its inner workings are only visible to the type system. This is because
> the relational system can only look inside predicates.

I don't care how it is implemented. As far as I'm concerned, a relation is a type, just as a bag is a type or a string, or name.

> Maybe there is some system of basic logic where lists are fundamental
> concepts but I'm not aware of it (and I'm not being facetious here).

English?
OK, I did take grad level logic courses once upon a time and I do recognize that a formal system of logic is important for querying the data, for example. But I use lists in English propositions often. Think of lists as connected by ANDs just as separate columns are except that they share the same type. What's so difficult about that? I guess the variable length of the list is one thing that could cause difficulty, but since it doesn't cause any difficulty in the environment in which I work, I suspect that any theorectical difficulty has work-arounds in the practical world.

> > and it is only if there is a question that requires a function to
> > look into the values of L would we then apply that function so that
> > we could ask:
> >
> > who has e-mail addresses at aol.com?
>
> yes, but that function is part of the type system and not the relational
> system.

again, the implementation of the model is of little interest to me as long as it works

> > Given that we can use SQL to GROUP data and if we can also UNGROUP
> > data, then where does the 2nd order predicate logic come in and why
> > is it a problem? Again, I really am ignorant on where we hit a wall
> > with first order predicate logic and what the problem is with
> > introducing additional logic to handle nesting and unnested as
> > needed. Thanks for your help. --dawn
>
> I'm struggling as well to understand exactly what it means in terms of
> relational database theory.
>
> Basically I think in second-order logic your "variables" such as L above
> are allowed to be predicates. Whereas in first-order logic they aren't,
> they are just allowed to be symbols with no structure of their own
> visible to the logical system.
>
> So that explains why Codd's relational theory can't have nested
> relations - because he chose to start with first order logic.
>
> The second question is why not start with second order logic?

I can understand why not to start there, but why not progress there?

> Well, its theory is a lot more complicated and you get stuff like
> Godel's theorems biting you with things like unprovability and
> incompleteness.

So what? Mathematicians have put up with Godel's theorem for quite some time and we haven't thrown out the Real numbers along with various operators because of it.

> Where I'm not totally clear is how much this has to do
> with infinities and whether the fact that databases are finite makes a
> difference, or whether their unboundedness is sufficient to cause
problems.

I don't know either. I would like to get my brain around this so I understand the arguements for eliminating lists in our data models.

> Maybe we need to cross-post to sci.logic for some expert knowledge of
> this kind of thing.

If you have the inclination to do so, that would be great. I'm too green on logic right now (and to think a professor once suggested I head towards a Ph.D. in the subject -- just one difference between my brain today and where it was in my 20's!) Thanks --dawn

> Paul.
Received on Fri Nov 05 2004 - 22:15:44 CET