DataBase Debunking

[Massimo Dentico]

Lately I was quite interested in practical application of the theory
of relations so I posted on #tunes channel on OpenProjects IRC networks
(logs here: http://tunes.org/~coreyr/) a couple of links:

- ap5, which extends Common Lisp with relations, operations on =
relations,
  consistency rules, and automation rules. Efficiency details are =
declared
  via annotations (http://ap5.com/)

- RelMiCS, Relational Methods in Computer Science
  (http://ist.unibw-muenchen.de/relmics/html/)

Apparently unrelated to this I was searching the web for a decent,
lean, free DBMS and then I "stumbled" upon this very interesting site,
DataBase Debunking (http://www.dbdebunk.com/).

Now I realize how enormously deceptive was my exposition to various
"explanations" (a better term could be _mystifications_) of the =
relational
data model and its pratical application.

A couple of excerpts from the site, EMAILS & NOTES section, (yes, Chris =
Date
contributes to the site), my note in curly brackets:

-------------------------------------------------------------------------=
-

08/24/2000 - Response to Article - "GREAT NEWS, THE RELATIONAL MODEL
  IS VERY MUCH ALIVE!" by Robert Stuart, with replies from C. J. Date
  and from Editor
 =20
  _The_ problem with relational databases is that nobody has ever seen
  one (in the commercial world, at any rate). We've all seen _SQL_
  databases, but _SQL databases aren't relational databases_. SQL itself
  suffers from all kinds of problems (of which NULLs and duplicates are
  two of the worst), but the single biggest overriding problem with SQL
  is that it doesn't support the relational model!

07/25/2000 - Response to Article - "WHAT DO YOU MEAN, POST-RELATIONAL?"
by Uriel Wittenberg, with replies from C. J. Date and from Editor

  - The logical and physical levels of the system must be kept clearly
    apart (though they often aren't). { Often (invariably?) a row in an
    R-table is equated with a *physical* record; "denormalization for
    performance": normalization is at the _logical_ level and has no
    impact on performance (physical level). If a DBMS perform poorly
    with relations in normal form then it is a bad implementation. See
    discussions on True Relational DBMS on the site }
 =20
  - Relations are n-dimensional, not two-dimensional. { R-tables are 2D
    but they merely *represents* reletions, they are not relations
    themselves }
 =20
  - Domains and relations are together both necessary and sufficient to
    represent absolutely any kind of data whatsoever.
 =20
  - The *one* good idea of object technology is support for a proper =
type
    system, and the relational model already includes such support
    { Domains in realational data model are ADTs (Abstract Data Types) }
 =20
  - Object technology includes several *bad* ideas (object IDs are one =
of
    the worst). Objects and a good model of type inheritance are
    incompatible.
 =20
  - Object technology is not a good basis on which to build databases in
    the classical sense of that term.
 =20
  - A true "object/relational" DBMS would be nothing more nor less than =
a
    true relational DBMS.

-------------------------------------------------------------------------=
-

Regards.

--
Massimo Dentico (MaD70)