Re: 1GB Tables as Classes, or Tables as Types, and all that refuted

From: Alfredo Novoa <>
Date: Thu, 09 Dec 2004 17:21:28 GMT
Message-ID: <>

On Thu, 9 Dec 2004 17:44:45 +0100, "Ja Lar" <ingen_at_mail.her> wrote:

>WHY is it "not possible to use a class to define a relation type" in the
>sense I'm quite sure Dawn refers to: In the same sense as we use an Entity
>in ER-modelling to define a relation (intentionally loosely speaking, you
>know ....)?

I asumed that she meant class in the same sense of scalar type.

>> No, it refers to the products that mix classes and tables.
>> You only have to review the first chapter of TTM to see that.
>TTM does appear to talk theory, not products.

They talk about the theory behind existing products.

>> It is about existing products, but I don't use such products.
>You don't use existing products :-)

I have the dislike to use several existing database products every day.

>Why shouldn't Dawn be able to see what you see without using such products,
>when you seems to be able to do it without using any products?

I saw the predicted practical problems making the experiment a few years ago.

>> Of course there is, but only if you use an OODBMS or an OR Mapper.
>So in a RDBMS there is no GB in equating class with relvar?

If you are using an RDBMS you can't equate class with relvar and if you are using a product that equates classes and relvars it is not an RDBMS. For instance with an RDBMS you can not do something like this:

var Customers real relation { Code Char, Name Char }

   key { Code };

var Orders real relation { Num Integer, Customer Customers }

    key { Num };

Here I am using a relvar as a type.

The 1GB consists in to design a product that allows something like this.

Regards Received on Thu Dec 09 2004 - 18:21:28 CET

Original text of this message