Re: The problem with denormalization.

"Craig Alexander Morrison" <reply_at_newsgroups.com> wrote in message news:41582599_at_212.67.96.135...
> One wonders if people do not know how to normalise data within the problem
> domain.
[snip]
> Usually people "denormalise" because they have attempted to normalise the
> real world for use in a business system.
>
Mostly as an accident of my own learning curve, I tend not to worry about normalization in the problem domain.

I adopt a three model approach.

The conceptual (or analysis) model describes the data in problem domain. ER model of the subject matter, and the data grouped into attributes with domains. Normalization is not an issue. There's a parallel issue, distinguishing between entities and attributes, and making sure that attributes are connected to the right entity. But it isn't normalization as such.

The logical design model, in which attributes are grouped into relations, and in which relationships are expressed as foreign keys.

The physical design model. using SQL tables and indexes, plus whatever lower layers (like TABLESPACE) the DBMS needs, and whatever process layers, like VIEWS, the application wants. I guess you can call this the physical model, but I think of the tables and indexes as being a reflection of the logical design.

The above is for a classical database. If I'm designing a star, the whole design is different, and even generally the analysis. Received on Mon Sep 27 2004 - 17:55:56 CEST