Re: The Theoretical Foundations of the Relational Model

On 26 Jun 2002 18:31:38 +0200, hidders_at_hcoss.uia.ac.be (Jan.Hidders) wrote:
>> So, if a product is going to support the three-layer model, they would
>> have two kinds of views, internal and external, and one would allow
>> sort and the other would not?
>
>Well, I would say that views are always part of (the mapping from the
>conceptual level to) the external level so there would be only one sort and
>it would allow ordering. But this placing of views in the external level is
>not uncontroversial.

"Conceptual level" is relation, "internal level" is b-trees and the like, and "external level" is PL/1 type definitions, per Date. So, where did you say we hang the sort? Of you, me, and Date, I don't think any two of us are talking about the same thing.

I was accustomed to conceptual/logical/physical as the three levels, all relational to one extent or another. Per Date, there's only one level that's relational, the conceptual.

Shrug. I continue to fail to see the point.

>Try Date's seventh edition, section 2.2 "The three levels of the
>architecture" (but earlier editions also have it).

Aha, as above.

> Actually most
>introductory works on databases will have something on it. If you are a
>member of ACM you could also try the original report:
>
> http://www.informatik.uni-trier.de/~ley/db/journals/sigmod/BurnsFJKMRRRT86.html

I'm a member, but haven't done the proper voodoo, apparently, cuz this just gets a reference header.

>> The three-level model *seems* like a good idea, and maybe it can be
>> salvaged, but the only encounters with it, or its advocates, I've ever
>> had have been massively unconvincing.
>
>The reason that I brought it up is because it helps to explain what it is
>that the RM exactly claims and how this claim is supported. If you
>understand that then you also may understand when the RM is not such a good
>idea.

Aha. Somehow we drifted into that from discussions on identity and sorts.

J. Received on Thu Jun 27 2002 - 01:09:27 CEST