Re: relationship in the database

In article <ampb90$f3e$1_at_sp15at20.hursley.ibm.com>, Paul Vernon <paul.vernon_at_ukk.ibmm.comm> wrote:
>>>In Date's words 'a thin layer on top of the basic relational model'
>
>>If you mean, can we translate an ER model to the relationa model,
>>then yes, of course.
>
>OK, but more importantly, do you agree that the RM model is more minimal
>than ER ?

I'm not sure that is such an easy answer. The structural part is simple, sure, but the complexities will return when you start specifying the constraints. And how complex is the structural part of an ER model? I would guess that the notion of a finite directed labeled graph is enough. Compare this to: "a function that maps a finite set of relation names to sets of tuples with the same domain, where a tuple is defined as a function that maps a finite set of column names to an atomic value".

>And that it is sufficient - what is possible in ER is also possible in the
>RM?

Agreed.

>I we agree with the above, then I come to my philosophical point that, to
>a large degree, I don't care if one model is more 'natural for people to
>think about' than another.
>
>For one that is a very subjective test, and even if a 'is more natural'
>test could be commonly agreed upon, it _still_ does not imply that the
>answer is to use a 'more natural' model rather than to improve education
>and toolings so that the people can think in the terms of the more pure
>model.

Why? Is pureness a goal in itself? If your data model is complex it will not get simpelere if you map it to the RM. I believe it is important that we keep in mind what the RM is for: a common ground to describe data that is viewed by many users in different ways, but at a high-enough abstraction level such that we can have our much-wanted data independence. My point was that the ER model is at a higher abstraction level, and so gives you better possibilities van data independence.

I should probably explain this a little more, but I have to get back to work.

• Jan Hidders