Re: The term "theory" as in "database theory"
Date: Thu, 01 Feb 2007 14:42:38 GMT
"dawn" <dawnwolthuis_at_gmail.com> wrote in message
> On Jan 27, 3:11 pm, "Marshall" <marshall.spi..._at_gmail.com> wrote:
> > On Jan 27, 12:53 pm, "dawn" <dawnwolth..._at_gmail.com> wrote:
> > > The false notion that Occam's razor (you can start with http://
> > > en.wikipedia.org/wiki/Occam's_Razor if you need more explanation) has
> > > anything to do with relational theory and the notion that because a
> > > relation is simpler than a graph, for example, that is proof that a
> > > relational model for data is better than a graphic model for data. --
> > If one is going to deny Occam's Razor, one is implicitly asserting
> > that, given two models of equal value, either:
> I suspect you are not catching my point on this. My point is that
> Occam's Razor is relevant to theories which seek to explain, but not
> to those which are effectively "how to" rules that we create based on
> a set of requirements. I see "database theory" as the latter.
I disagree, on several points.
First, Occam's razor is useful to compare two "theories of data". If the two theories are equally valid over the same scope, then the simplest one is to be preferred, according to Occam's razor. If all other things are not equal, then preference is a unidimensional domain projected from a multidimensional space of values. Simplicity is one of the dimensions.
The same argument, using Occam's razor, is equally applicaple to two database designs. If they are equally valid over the same scope, then the simplest one is to be preferred. However, you have other measures of "goodness" operating here, besides simplicity. One is performance. Another one is revisability. Another one is ease of administration. etc.
The same principle, using Occam's razor, is relevant to the comparison of two sets of database design principles.
> We can still talk about simplicity as a heuristic, and from my
> perspective it is overall simplicity that is desired, not the simplest
> code, the simplest model, etc, but the simplest overall solution that
> meets the requirements for the software.
> Make sense or not? Thanks. --dawn
this last doesn't make sense. Received on Thu Feb 01 2007 - 15:42:38 CET