Re: database systems and organizational intelligence

"Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message news:c97uml$rd0$1_at_news.netins.net...
> I paid for it and I can give my impression (which you won't like) -- that
> relational theorists have finally concluded that 1NF as we have learned it
> for a few decades now has no mathematical backing in relational theory
(but,
> of course, I knew that;-). So, they have to extend the theory now to
> provide for nested relations,

It's not an extension. It's simply another type. It's actually a type-generator (like a list or tuple or other collection) for which there's already a well-defined model. That means you have lots of useful operators, and don't have to write as much code. However, the power of a type-generator comes at a cost, which is why Date and Pascal caution against using it when other types (including your own) will do.

> which, of course, have been part of every
> software application that is not relational since I don't know when. Then
> Pascal's document (which I also paid for) comes along and says "yes, but
> here are the reasons why you don't want to use nested relations even
though
> they are mathematically sound" thus proving that even if a theory is
> mathematically sound (e.g. relational theory), that doesn't mean it has
> enough practical value to employ it.

That's true, but Date gave a number of solid examples of when you would want to use it: some family relationships, groupings, and the elimination of outer join, which has problems. All that makes the theory more consistent, and also allows you to use the existing algebra in new ways, without mangling it. e.g. for reports, which could identify the relational type and break into a sub-report... etc.

> smiles (you are smiling, right?) --dawn

Well, my teeth are clenched. Ah, just kidding...

• erk