Re: In an RDBMS, what does "Data" mean?

"x" <x-false_at_yahoo.com> wrote in message news:40d000b4_at_post.usenet.com...
> > A star schema serves different purposes than a fully normalized design.
> > Depending on what you are trying to do, it could be the right way to
go,
> > or it could be an unfortunate design decision.
>
> As I understood Codd's papers, a fully normalized design was not intended
> for users/applications schema.

That's how I understood Codd's papers as well. Unfortunately, some people have made normalization and the RDM into some kind of quasi religious mumbo jumbo by which the true believers can transubstantiate garbage into holy data. And the people who know the secret handshake, well they are the higher order Freemasons.

This is doubly unfortunate because it disguises the actual value of normalization and the RDM, which is quite large. But it isn't the silver bullet, just as avoiding "GOTO" wasn't the silver bullet and encapsulating local static variables wasn't the silver bullet.

>
> > A star schema is still in 1NF, so there's no issue there. Many of the
> > dimension tables have a simple key, so they are automatically in 2NF.
The
> > fact tables typically aren't even in 2NF. But for 3NF and beyond,
there
> is
> > simply no attempt to normalize at all. You pay the price in redundancy,
> > and you pay the price in difficulty to load.
>
> One could design a star schema as a user/application schema.
> There will be a drop in performance ?
> Why ?

When updates happen to the dimension tables, it may be that multiple updates are required and that slows things down.

Some joins may be slower, because of the key structure of the fact tables. I haven't done enough playing around with fact table indexes to know everything I ought to know about this.

>
> > But you gain enormous power and simplicity when you want to take the
same
> > data, and slice it and dice it every which way, or drill down for more
> > detail, etc. etc. There is even a tool, from the vendor of SQR (I
> forget
> > who) that does all kinds of OLAP games, but has no MDDB of it's own.
> > Instead, you plug it into your star schema, and you're ready to rock
and
> > roll!

>
> I've heard that MD model is (mathematically) equivalent to relational
model,
> but I cannot remember where.

I wouldn't know. In practice, they sure feel different.

> I also heard that a star schema is a properly designed schema.

There's more than one way to skin a cat. For a star schema (or any other design), there will be people who hail it as the holy grail of design and others who dismiss it as the dumbest thing they have ever seen. You can't please everybody.

Heresy is nearly always a necessary part of advancement or reform. Martin Luther, Galileo, Frank Lloyd Wright, and e.e. cummings were all denounced as heretics by people who liked things the way they were before. But most heresies are failures, just like most mutations are failures.

And the present age is so in love with heresy that we teach kindergarten kids to color outside the lines before we teach them how to color within the lines. Nobody thinks inside the box anymore. And some don't think either inside or outside the box.

It's a puzzlement. Received on Wed Jun 16 2004 - 14:37:03 CEST