Fw: [thelist] SQL query help

rudy r937 at interlog.com
Fri Jan 4 15:30:39 CST 2002

forwarding a message from myself that seems to have disappeared into some
internet black hole yesterday

-----Original Message-----
From: rudy <r937 at interlog.com>
To: thelist at lists.evolt.org <thelist at lists.evolt.org>
Date: January 3, 2002 09:43
Subject: Re: [thelist] SQL query help

> I hate to get drawn into debates, but the real myth is
> that a fully normalized relational database is the
> only proper construction.

hi joel

there was no debate, and there are plenty of myths when it comes to
normalization (see fabian pascal's article "Denormalization For
Performance" - Et Tu Academia? http://www.firstsql.com/dbdebunk/fp5a.htm)

the myth i was referring to was the common tendency for people to consider
denormalization at the first suspicion of poor performance

as you said yesterday, this is probably not a good move unless you are
dealing with tables of millions of rows and result sets of thousands, and
have exhausted all indexing strategies

> I always design normalized structures initially, and I
> am a very strong proponent of surrogate keys.

rock solid approach

i am amazed that so many people, new to database design, use surrogate keys
so easily -- it must be that surrogate keys are being taught in all the
intro courses, without that specific name, because many people don't know
what a primary key is, or think an autonumber is automatically the primary

> Obviously one uses caution before resorting to data
> redundancy

i would go further and suggest using redundancy only as a last resort


More information about the thelist mailing list