Re: QUESTION: List array, graph or network model support DB
Date: Sat, 14 Dec 2002 11:17:31 -0800
David Cressey wrote:
>>if you mean by "ideal" that it runs on Unix and crashes all the time and
>>needs a bazillion DBA's to keep them running and you want to constantly
>>recover your database and your data files, then you can have ideal.
> A little background on my original comment might be in order. I don't tend
> to use the term "ideal" myself, much.
> I was referring to a comment made fairly frequently in this forum, to the
> effect that "A commercial Relational Databse system has never been built."
> These people exclude Oracle, SQL Server, DB2, Informix, Interbase, yada
> yada, because all of them fail, in one way or another to live up to the
> "ideal" of a truly relational system. I have a hard time with such
> terminological rigidity, myself. One can say that all those products
> aren't perfect relational products, but one shouldn't, in my view, say
> that they "aren't even relational".
Why do you think that they are "relational" ? Do they operate on relations ? I don't think so. If their primary business is not to operate on relations but on bags of rows, calling them relational is misleading.
Just like ODBMS are often database construction kits or persistence libraries, SQL DBMSes are a real DBMS (they do provide transactions, recovery, concurrency control, some data integrity) + a *relational construction kit*. Meaning that by a skillful use of SQL one can come somewhere close to a relational database.
But the complexity is left on the user to shoulder, and it is very difficult to stretch SQL so that you are still in the realm of relational model. And guess what: most users don't and most users suffer as a consequence.
It's even worse than that : very often product documentation and books sponsored by the vendors (Oracle press: anyone there ?) simply lie to the users by defining relational model in the most ridiculous terms. Actually they screwed up their products, they built a multi-billion dollars industry by taking agressive shortcuts on the implementation side and transfering the complexity to the user and now they try to lie and cheat by proclaiming their version of "relational" (not long ago the auto industry maintained seat belts and airbags were unnecessarily expensive and not needed).
Costin Cozianu Received on Sat Dec 14 2002 - 20:17:31 CET