Re: Design question

From: Todd Hoff <tmh_at_darla.asd.sgi.com>
Date: Fri, 14 Jan 1994 21:18:10 GMT
Message-ID: <CJn1uA.C7t_at_odin.corp.sgi.com>


In article <2h6kk0$b0c_at_athos.cc.bellcore.com>, derek_at_candyland.gamekeeper.bellcore.com (C. Derek Fields) writes:

|> To state the question in a more abstract fashion: If, in a M:1
|> relationship, we know
|> that M is a constant value, what are the advantages/disadvantages of
|> normalization.
|> It is my hope that by asking modeling questions, I can stimulate people
|> to ask their
|> own modeling questions and participate in a modeling dialogue.
|>

A couple of points:
1. Most of us still have the burn marks from the times when we have

   beleived someone when they said "I'm sure, it will only ever in the    whole history of the world be X." Experience and murphy say: if it    can be 2 then someone wil want it to be 3, then 4, etc.. 2. If performance is major alpha critical then denormalize, where is the choice?
3. Using variable like can1 and can2 yields a lot of code like:

      if (can1 == 2 || or can3 == 4) then    Adding another can breaks ALL of this code! Using sets solves this problem.

-- 
Todd Hoff
Received on Fri Jan 14 1994 - 22:18:10 CET

Original text of this message