Re: Codd provided appropriate mathematics ... (was Re: Relational and MV (response to "foundations of relational theory"))

"mountain man" <hobbit_at_southern_seaweed.com.op> wrote in message news:p8Z_b.75518$Wa.20615_at_news-server.bigpond.net.au...
> "Marshall Spight" <mspight_at_dnai.com> wrote in message
> news:p5V_b.117800$jk2.509873_at_attbi_s53...
> > >
> > > I doubt whether one could in fact benchmark the relational *model*,
> > > but rather only an implementation instance of the model. Thus I'd not
> > > expect there to ever be any empiricism in this area.
> >
> > I think Dawn is referring (in the second part) to the possibility of
> > testing whether software development is more cost-effective using
> > one data model vs. another. This sort of thing is quite difficult, but
> > it is certainly possible. I have, for example, read some interesting
> > papers comparing programmer productivity in different programming
> > languages.
>
> How do the results of these papers stack up the
> programming languages in terms of productivity?

The general takeaway is that programmer productivity is quite similar across languages as measured in lines of code, but that amount accomplished per line of code varies with language. This argues for terse languages. I don't think enough of these studies have been done to consider this proven, though.

As a related aside, I observe that SQL is the COBOL of query languages; it spells everything out to an appalling degree. I can imagine other query languages that would be much more concise.

> > As to the first part, I would assert that right and wrong are not
> > qualities of a model. Rather, models are said to be more or less
> > useful. I think Dawn has a valid point in critiquing the RDBMS
> > world's tendency towards saying that the RDM is the only
> > mathematically sound model.
>
> When you look under the bonnet of the RDBMS software
> available today there are impressive features not previously
> available to 1980 DBMS architecture.
>
> There is always a second issue, namely, that irrespective of
> the (R) in (R)DBMS, the actual implementation is always one
> step removed from the data model used.
>
> A good DBA team implementing a non R-DBMS can achieve
> a better result than an inexperienced DBA team implementing
> a supposedly "superior" RDBMS. It is this issue I refered to
> above.

Okay.

Marshall Received on Wed Feb 25 2004 - 17:02:20 CET