Re: Dawn doesn't like 1NF

"Alfredo Novoa" <alfredo_at_ncs.es> wrote in message news:4163d4ef.5438328_at_news.wanadoo.es...
> On Tue, 05 Oct 2004 16:37:56 GMT, "Marshall Spight" <mspight_at_dnai.com>
> wrote:
>
> >Time to refresh my memory. Remind me again what it is you
> >don't like about 1NF. Is it simply that one can't have relations
> >as attributes? (Or lists or whatever, although some SQL engines
> >allow lists as attributes.)
>
> You can have relations as tuple attribute values in 1NF.

I have a half-written response to the original question, but I'll jump in on this anyway. When I mention 1NF, I mean 1NF as it was defined until a few years ago and the 1NF that SQL-92 (and, therefore ODBC) understands, and the 1NF that most RDBMS users put their data into. So, I am very pleased that the industry has gotten back to the concept that a value could legitimately be a list. This is not well-known or well-taught as yet, although it is only with relational theory that the flawed notion of 1NF was brought into the database picture. Thankfully gains are being made, but it will take a long, long time before we fix our industry -- the 1NF craze (as previously defined) has been very costly.

> >If you could have a column that had a relation type, or a list
> >type, and these could have relation types or list types in turn,
> >recursively, is that what you want?
>
> What she wants is to manually transverse hierarchies using procedural
> code in the ancient Pick way.

Wrong, bucko, I mean, Alfredo ;-)
One can specify graph paths in a variety of ways -- the specification can be metadata and not procedural code -- I am not referring to any implementation of the theory. I'm saying that the same theory humans use to navigate roads and the web can be used to navigate data. Navigation is NOT bad and does have a mathematical theory behind it as well. When working with data, we can use the mathematics of sets, predicate logic, relations, functions, AND graphs (di-graphs in this case). Why eliminate some good strategies for working with data for the sake of sticking only with set theory? There are good reasons to "view" the data through a particular node without referring to the entire set of such data. More later. --dawn

>
> Regards
>
Received on Wed Oct 06 2004 - 16:16:07 CEST