Re: One to One relationships

From: stu <smcgouga_at_nospam.co.uk>
Date: Fri, 25 Apr 2003 09:07:22 +0100
Message-ID: <b8aqfl$snq$1$8300dec7_at_news.demon.co.uk>


> Now the *relationship* Locations can have attributes -- first year of
> residence or whatever makes sense.

Ok so now my understanding is that if your model has a 1:1 which changes over time we could introduce a relationship that has a from and thru date. And what you seem to be saying, Celko, is that if the 2 items in the relationship exist separately in the real world (can one exist without the other) then they should be separate tables.

Obviously if you have millions of people and millions of houses your join from people to Locations to Houses is going to take a considerable amount of time compared to holding house as an attribute of person. Comments? And no I don't think we should just put all data in a massive flat file structure but surely for a system with few updates and lots of queries this would be advantageous.

My uni lecturer once said do not do physical design (i.e. do not denormalise for speed). Let advances in hardware benefit your correctly designed DB. But if I design databases like this and my competition designs faster DBs who gets the work? I know this is a theory NG but if there are faster, different methods to do things we should look at those right?

Cheers
Stu Received on Fri Apr 25 2003 - 10:07:22 CEST

Original text of this message