MORE ON THE LOGICAL-PHYSICAL DISTINCTION
with Fabian Pascal

 

 

 

From: CB

To: Editor

Date: 09/16/2003

 

In your writings, you have made the point that a data model is something that's logical (e.g., orthogonal to its actual physical implementation).

 

Does that mean, for example, that a DBMS could be truly relational if, under the hood, it is physically built upon such traditional programming constructs as linked lists, pointers, and arrays?

 

And even if so - are you aware of a better way to develop a TRDBMS without using such constructs?

 

 

From: Fabian Pascal

To: CB

 

A data model is logical by definition. In the case of the relational model this is made even clearer by the fact that RM is nothing but logic applied to database management. How can logic say anything about physical implementation? It's absurd.

 

An implementer has complete freedom to use whatever physical means at his disposal, as long as he does not expose them to users in applications, and does not make logical features dependent on specific physical details.

 

The best approach to TRDBMS implementation we know of is the TransRelational(TM) Model, which does use logical pointers at the system level, but does not expos them, nor any physical details to users and applications.

 

Chris Date gave a one-day seminar on it at UCLA in September. We hope to publish on the subject if and when we are permitted to do so.

 

 

Posted 11/14/03

 

 

 

[ABOUT] [QUOTES] [LINKS]