Re: Why bother with Logical data model?

"Brian Selzer" <brian_at_selzer-software.com> wrote in message news:2tjCg.4506$o27.308_at_newssvr21.news.prodigy.com...
> Why? Because it saves time--perhaps not initially, but definitely in the
> long run. In most of the databases I've worked with there is a disparity
> between the logical model and the physical model. The logical model
should
> at a minimum be in BCNF, but if possible, 5NF, but that is not always true
> in the physical model. Sometimes tables must be split for valid business
> reasons, such as security. Sometimes, though seldom, a denormalized table
> performs better--at least for the particular application where it is most
> used. The logical model should remain independent of such considerations,
> so that you can focus on the structure of the data that is to be stored

The above is a very, very good point. It deserves careful attention by most of the readers of this newsgroup.

In almost every case I've dealt with, 3NF was "good enough" at implementation time. In the case of star schema design, 1NF was "good enough", but that's a digression. Anyway, the mental effort of generating a fully normalized model was worth it as a cross check on the data analysis! And that sometimes uncovers flaws in the requirements themselves!

IIRC, you've made this point elsewhere: there are really three levels of modeling: Conceptual, Logical, and Physical.

The above outlines the difference between logical and physical nicely. The difference between conceptual and logical is, IMO, just as important. Received on Wed Aug 09 2006 - 13:33:41 CEST