Re: The Theoretical Foundations of the Relational Model

In article <3d21a4fc$0$223$4d4ebb8e_at_read-nat.news.nl.uu.net>, Jimmy Venema <jimmy.venema_at_philips.com> wrote:
>I always hear about the well founded mathematics of the relational model,
>but if I look at the basics I only see stripped OO models.

Stripped models? Interesting choice of words. :-) Anyway, this simplicity is by many seen as one of its strengths, not its weaknesses. Do you see any good arguments to make things more complicated?

>From a OO point of view a relational database implements its relations only
>unidirectional, it only has what we call a backward reference.

A foreign key doesn't have a direction in the sense that it tells you in what order you should do you joins or that you should go from relation A to relation B and not vice versa.

>In a OO model you can have a forward unidirectional relation a backward or
>a bidirectional relation. Most of the time people are not aware of this and
>during implementation they make a selection. In order to improve forward
>navigation or to phrase it differently, improve some typical kind of SELECT
>use, additional (binary) index trees are used. All this tree stuff happens
>behind the scene, but it is there to get decent performance.

So it walks like a duck, and it talks like a duck, and it is often implemented as a duck, but *conceptually* it is not really a duck. Is that what you are trying to tell us? :-) But seriously, you are certainly right that strictly speaking also OO models do not tell you in which direction you should follow a relationship.

>So if I look at the table/colomn level and the class/member level I must
>say that the relational model is a subset of the OO model.

Sure. You can simulate RM in an OO model and you can simulate the data part of an OO model in RM. I might also claim that many OO models are in some sense very limited relational models because they only have binary relationships. What's your point?

>[...] The problem is that there are additional constructs
>possible and for that part there is not (yet) a well founded set of
>mathematics.

Actually there have been quite some formal mathematical descriptions of OO data models and also of OR data models. The problem is more that it is not clear which one to choose and why.

• Jan Hidders