Re: The Practical Benefits of the Relational Model

In article <ankj7e$13a6$1_at_sp15at20.hursley.ibm.com>, Paul Vernon <paul.vernon_at_ukk.ibmm.comm> wrote:
>>>> God forbid nesting/unnesting!?
>>>
>>>Fully allowed.
>
>>It is? So how is the 1NF defined then in the relational model? Or is there
>>now a new "*the* relational model". :-)
>
>Scientists don't currently seem to work that way. I.e. AFAIK there is not a
>single body that administers *the* relational model, unlike say for SQL.

Sure, but a good understanding of what the 1NF exactly is seems rather essential to me for the relational model.

>Nor am I aware of any body centrally defining the Theory of Evolution, or
>even Gravity.

I don't think that's comparable. The relational model is not a descriptive theory such as those that are found in science, but rather an engineering model that tells us how to design something and incorporates a lot of experience from the past. Note that I don't want to suggest that this is lesser than a scientific theory, but the two are different.

>Why should Relation values be explicitly the SINGLE datatype out of the
>universe of all possible database that are disallowed from being attribute
>values?
>
>And besides that, they are darn useful in cases like the System Catalog.

Ok. But why then disallow the operation of unnesting? If I want to associate A's with a set of B's then I have two options:

  R(A,C) where C is a set of B's

and

  R(A,C) where C is an identifier for a set of B's   R'(C,B) and here the set for each identifier is described.

I would say these are more or less similar, although in the second option I have to invent some kind of identifier. But in the second option I can express the unnesting and in the first case D&D would not allow the unnesting. Why not? What is the big problem here?

• Jan Hidders

PS. Sorry for the late reply, I'm moving from one appartment to another at

    the moment, and therefore not on-line in the weekends. Received on Mon Oct 07 2002 - 16:39:45 CEST