Re: Daatabase

On Wed, 17 Aug 2005 08:32:32 -0700, lou_nyc interested us by writing:

> Normalization should be derived from the business rules. In other
> words, weather you normalized or
> denomalized depends sorely on your businesss. However, that is not the
> case when it comes to
> a Data Warehouse implementation.

Are you implying that Warehouses have no business rules, or simply don't follow business rules? ;-)

If you reread my comment, I am simply stating that the final state (normalized or denormalized) should be determined after appropriate analysis - and that a proper analysis usually implies normalization first, followed by *intelligent* denormalization to achieve specific goals.

(My point is too often I've seen that analysts don't analyze and then use the "it's denormalized" excuse to cover their failure.)  

I totally disagree that report generation and analytics can avoid similar detailed analysis.

I do agree that implementing that report and analytics often can best be resolved by dimensional methods. And I further agree that in those cases people can use shortcuts like star and snowflake designs to go directly to implementation. But (IMO) that should be left to people who have experience using such shortcuts. (I have a similar problem with the thoughtless 'design by pattern' often seen these days in the Java world.)

-- 
Hans Forbrich                           
Canada-wide Oracle training and consulting
mailto: Fuzzy.GreyBeard_at_gmail.com   
*** I no longer assist with top-posted newsgroup queries ***

PS: Since the above discussion, and many counterarguments, falls
dangerously close to religion, I leave the discussion at this point. <g>