Data management principles (formerly Why using "Group By")

"Mikito Harakiri" <mikharakiri_at_ywho.com> wrote in message news:8M7ca.21$wV5.64_at_news.oracle.com...
> "Bob Badour" <bbadour_at_golden.net> wrote in message
> news:bI7ca.2$7Z2.139401_at_mantis.golden.net...
> > "Mikito Harakiri" <mikharakiri_at_ywho.com> wrote in message
> > Since I first observed the problem with sybase in 1994 four years before
> > Date's article, I doubt the problem has anything to do with that
> particular
> > article. And in any case, the result in sybase is not the same as the
> > interpretation you suggested.
>
> OK.
>
> > Further, having just re-read the article, I do not see where Date
proposed
> > any such thing.
>
> A transformation of "group by" to scalar subquery is certinly there,
> although, I'm uncertain if the verb "proposed" could be really applied to
> what he said.

Yes, a transformation is there, but it is not the same transformation you attributed to him.

> > > What is logical identity, that you refer to so often?
> >
> > Are you saying that you are unfamiliar with Codd's "12 Rules"? Or are
you
> > saying you are unfamiliar with the concept of identity? Or are you
saying
> > you are unfamiliar with the logical level of discourse?
>
> Ah, those 12 apostols! I've read that article maybe twice, but those rules
> are constantly slipping my mind. Unlike really exciting things like
> "principle of equivalence of mass and inertia". Either, there is something
> wrong with me, or those rules aren't really worth much. Could somebody be
so
> kind to reduce the number and formulate one or two really fundamental
> principles instead? Like relational closure, for example?

Logical identity would certainly top the short list of necessary data management principles. Logical identity really is fundamental: It is hard to make meaningful statements about a thing if you cannot even identify what it is.

(In terms of Codd's 12 Rules, logical identity is assured by the Guaranteed Access Rule.)

A data management system should allow users to manage data without distracting them by requiring them to manage inconsequentials.

Other principles? Human knowledge changes rapidly but human nature changes slowly. This leads to what Steve Maguire would call the "Candy Machine" design principle. Design to bias human behaviour to desired outcomes and if possible to prevent undesired outcomes. This would apply to the design of data management languages in particular.

Closely related to the above is KISS.

And closely related to both of the above is the Principle of Cautious design.

Anyone have any others they would like to nominate? Received on Fri Mar 14 2003 - 04:27:20 CET