Re: object algebra

From: Eric Kaun <ekaun_at_yahoo.com>
Date: Wed, 03 Mar 2004 13:17:45 GMT
Message-ID: <ZVk1c.55377$JQ6.15645_at_newssvr33.news.prodigy.com>


"Drago Ganic" <drago.ganic_at_in2.hr> wrote in message news:c1qvlo$agh$1_at_sunce.iskon.hr...
> "Neo" <neo55592_at_hotmail.com> wrote in message
> news:4b45d3ad.0402242201.72ab2c26_at_posting.google.com...
> > Is "Not Applicable" a color?
>
> This question is really a good one and was not answered well !!
>
> Domain should be from the problem space (semantics). When we talk about
eye
> colors there are no N/A or UNK eye colors in the world. There are just
> "people" (tuples) who don't have eyes and therefore don't have eye colors,
> or people who have eyes but we don't know the color of them. So I thing we
> have to leave the domain intact.

What we have to do is model data. No model can ever match the complexity of the real world. So we must simplify, and represent things as simplistically as possible (but no simpler). Since we can define the domain and its operations, we're free to do that as we wish. We're also free to define what it means to "mix" eye colors (my answer: you don't), and to define what it means when I compare the eye color of Mary (N/A) with the eye color of Bill (UNK). The point is this: NULLs (NOT missing information, but NULLs specifically) force us into overloading and awkwardness, not to mention non-closure, confusing overuse of NULL to mean different things, etc. etc. So don't use them. Say what you mean for each domain. Received on Wed Mar 03 2004 - 14:17:45 CET

Original text of this message