Re: 4 the Faq: Strengths and Weaknesses of Data Models

From: Laconic2 <laconic2_at_comcast.net>
Date: Fri, 15 Oct 2004 12:26:41 -0400
Message-ID: <rqydnTZelajNZ_LcRVn-iw_at_comcast.com>


"Marshall Spight" <mspight_at_dnai.com> wrote in message news:BSRbd.252104$D%.145867_at_attbi_s51...

> > It was possible to define a "relation" that would violate 1NF.
>
> Argh! You're going to drop something like that and just leave it with
> no details?! Please be specific. You might have noticed 1NF gets
> debated here every once in a while. :-) Tell us more.

There's really no need for specifics. It's the same issues that have been discussed endlessly in here, about "what wrong with tables". Things like allowing duplicates unless you declare a primary key constraint.

> > They had a datatype called "DBKEY". This was really a pointer. This
meant
> > that you could build your own network or hierarchical database
underneath
> > the relational database if you wanted to. DBKEYs stored in records
suffer
> > from all the problems with the graph data model that are outlined in the
> > theory. But they were blazingly fast!
>
> Hrgmmmph.

I never used this feature. Basically, a record that has a DBKEY pointing to it is "pinned", for the same reason a web page is pinned once there's a link to it. My reaction was: "if you want to play CODASYL, go get yourself a CODASYL DBMS."
I'm even inclined to agree with relational bigots in declaring this anathema.

>
>
> > Their DML was really not a PL. In order to do "real programming", you
> > needed a "real programming language".
>
> I guess that's the same choice SQL makes. Every query halts, but you
> can't express everything you might like to.

Yeah. SQL is really a DML and a DDL. It's really not a programming language. Although it's being dragged in that direction, willy nilly, by the same people who wanted to call it a "user language" years ago. This happens over and over. There was an article in Datamation sometime around 1970 with the title "The next 700 programming lanaguages" that used this idea as a starting point.

>
> I'm interested in more details if you have them.
>
Well, I took the Rdb Internals course in 1994. And I taugh a lot of lower level courses before that. But my memory is really fading on this subject. I'm not sure I can get it back.

I remember doing a point by point comparison between Oracle and Rdb sometime around 1995. I was quite surprised when the final score favored Oracle, by one point out of about 35. To make a long story short, Oracle was more "programmer friendly" but Rdb was more "DBA friendly." Received on Fri Oct 15 2004 - 18:26:41 CEST

Original text of this message