Re: OO and rdbms

From: Clayton Weaver <cgweav_at_eskimo.com>
Date: 2000/04/13
Message-ID: <Pine.SUN.3.96.1000413031121.17049C-100000_at_eskimo.com>#1/1


There was a reasonably intelligent thread on fitting OO application languages to rdbms backends in Usenet newsgroup comp.object in 1998. Search Deja News for "Subject: OO model on top of relational DB".

There was another one just recently (I have a post from Feb 2000), same newsgroup: "Subject: Isolating the database".

These aren't analyses of "object oriented databases". What is that exactly? I doubt that it is well defined in the abstract, although there are commercial apps like Gemstone that define their own model of what integrating OO and dbms systems amounts to.

OOdbms is not inherently "better" than rdbms or vice versa. Which is more appropriate depends on the usual structure of the data must be managed.

If the data structure naturally fits an hierarchical or semantic database model better than a relational model, using an OOdbms can automate some behavior that would have to be managed in application space if storing the same data in an rdbms. One presumes that using an OOdbms is less error prone for applications that must manage such behavior and data than doing it anew in ad hoc fashion for every application with similar data structure characteristics.

Do a web search for Gemstone or Opal and "database & object-oriented", you'll probably find some case studies that indicate what kind of data an OOdbms is considered useful for (I dimly remember skimming some documents on a Boeing research project or something like that which used one of these OOdbms to manage the data from simulation and compare with real-time sampling of physical phenomena, IIRC).

This page has possibly useful references on update anomalies and OO inheritance:

  Resolving Inheritance Abnormality in an Object-Oriented Paradigm   

  <http://www.cs.fiu.edu/scspage/professor/Sun/research6.html>

Hth,

Clayton Weaver
<mailto:cgweav_at_eskimo.com>

(Seattle)

"Everybody's ignorant, just in different subjects." Will Rogers

-- 

Clayton Weaver

<mailto:cgweav_at_eskimo.com>
(Seattle)
Received on Thu Apr 13 2000 - 00:00:00 CEST

Original text of this message