Re: Resiliency To New Data Requirements

Neo wrote:
> > The most general method of representing things is to use bits.
> > All that remains is the matter of filling in the details.
>
> If this is true, why don't data models (ie hierarchal, relational, etc)
> mention bits?

Because they've filled in the details.

> > Some of the first details that should be filled in would be a
> > specification of what exactly "generality" means with regards
> > to "representing" "things." Until there *is* such a specification,
> > I assert that the particular task you are trying to solve is unspecified.
>
> Would you consider the relational method of representing things to be
> more general than the hierarchal method? If so, why?

Yes, I would. Hierarchies only allow one to use a single organizational scheme; relations allow as many as are necessary. In fact, if you follow the normal forms, the organization scheme is complete abstracted from the structure of the data itself and becomes an attribute of particular queries instead.

Another difference: hierarchies have only the single way of indicating a relationship between two kinds of things, which is placing them in adjacent levels in the hierarchy ("containment".) This is sufficient for modelling many-to-one relationships, but fails at many-to-many relationships. Relations have no trouble with many-to-many relationships.
In fact, it was exactly this surprising quality of relations that pierecd
my OOP-smug-shield and made me pay attention to them.

Marshall Received on Mon Aug 07 2006 - 03:39:01 CEST