Richard Heyes wrote:
>> I do not agree that creating a database which is normalised to3NF is a
>> waste of time.
>
> It isn't always, but it is sometimes. When time is a (significant)
> factor, getting something up and running (which has acceptable
> performance) may be more impotant than creating a technically perfect
> solution. In fact creating something that is technically perfect is
> often just a pipe dream for programmers.
>
> > On the contrary, a totally un-normalised database is nothing but a
>> problem waiting to bite you in the a**e.
>
> So you can:
>
> a) Create something that gets you to market as fast as possible that is
> "good enough".
> b) Optimise/adjust the structure later.
>
> IME though, b) rarely happens.
>
> > Computer systems have a habit of
>> growing over time
>
> Really?
>
>> ...and if you don't follow the rules of normalisation your database
>> will end up as the biggest bottleneck.
>
> Granted it's more likely, but not a given. You just need developers who
> have discipline, oh and a good memory helps.
>
>> Anyone who doesn't know how to reach 3NF shouldn't be designing
>> databases.
>
> Rubbish. It helps, in particular for how you can optimise you structure
> without duplicating data (too much), but shouldn't be a requirement.
>

Me personally I've always found it very productive to take a few hours
before I begin coding a project, to roughly flow-chart the basics of the
application, and then layout the db on paper to get a graphical view of
the tables I'll need, how they relate or don't relate to one another.
That way when I do actually create the db I'm usually at 3NF.

--

Mark
-------------------------
the rule of law is good, however the rule of tyrants just plain sucks!
Real Tax Reform begins with getting rid of the IRS.
==============================================
Powered by CentOS5 (RHEL5)