Re: Testing Various Data Models?

"Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message news:c2m660$o1g$1_at_news.netins.net...
> "Marshall Spight" <mspight_at_dnai.com> wrote in message
> news:GQw3c.222763$jk2.857968_at_attbi_s53...
> > "Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message
> news:c2l1dq$2q3$1_at_news.netins.net...
> > >
> > > I would very much like to see such a competition because my intuition,
> > > without sufficient scientifically gathered emperical data, is that the
> > > relational model ought to be retired.
> >
> > That's kind of like saying set theory ought to be retired.

>

Well... there's only one, and it's a little young for retirement.

> However, a theory about mass
> murderers that isn't the best at solving such crimes is not the theory I
> would want to employ.

Computing depends on many theories, just as criminal investigation depends on forensics (which incorporates biology, chemistry, physics in general, fluid dynamics, etc. etc. etc.), psychology, sociology, etc. etc. etc. It's a complex discipline which relies on the stability and predictability of underlying science to work properly.

Now, if Zippy the Wonder Psychic comes along and is able to solve crime after crime using dowsing rods and crystal balls, that doesn't really speak to the repeatability of the "process." If Zippy (and no, I'm not comparing Zippy to you) arranges a competition of psychics against legitimate investigators, and the psychics happen to win, that still suggests very little. There needs to be an element of fundamental scientific truth underlying such things. Even investigators with good hunches need to substantiate their guesses, and they use science for that.

So this competition you propose would be very difficult to orchestrate in a way that properly measures anything, and you'd still need to somehow explain a success in terms of... well, something basic and understood like math and science.

> So, I think is not a huge stretch to suggest we
> retire the relational model for the purpose of modeling data to be stored
> and retrieved in databases.

Many have suggested it. All are wrong, for reasons spelled out many places, even in these newgroups. The transformations done in computing are for the most part straightforward, at least in business information systems, and demand a solid foundation, not a practice founded on inexplicable theory and needless complexity.

As an aside, another reading suggestion for all: Problem Frames, by Michael Jackson (the S/W engineer, not the pop star or the beer expert). An excellent application of domains and predicates to problem analysis rather than data modeling.

• Eric