[Nottingham] TALK: Relational databases, SQL and PostgreSQL.

Michael Erskine msemtd at yahoo.co.uk
Mon Jan 15 10:18:36 GMT 2007


On Monday 15 January 2007 09:49, Ovid wrote:
> A cardinal rule of software development:  optimize for *correctness*.
> Then, and only then, do you optimize for performance.

Yes, (almost) always true.

I was thinking of a particular design I'd encountered that had text book 
correctness and the denormalisation of which solved a number of problems in 
an expedient manner.

> Denormalization may be an option, but you have to have hard-numbers to
> back it up.  Otherwise, you're just guessing and this leads to more
> problems down the road.

The proof of the pudding was in the eating; and what a sweet pudding it was 
for very little additional effort. The database was designed by an outside 
agency and worked rather well until all and sundry wrote their queries which 
drove it into the ground. In no position to rewrite the queries of other 
contractors, I tested a few tweaks to help stop the deadlocks and went with 
whatever worked. There's more than one way to do it: whilst correct, the text 
book can't tell you the ways that are not known at time of writing. I like to 
go with a hunch: if I have a gut feeling that something's going to work I'll 
pursue that with more vigour than a dictum handed down from on high!

Regards,
Michael Erskine.
(only exercising the keyboard (or perhaps exorcising!). Please don't criticise 
my overuse of the colon :) )

-- 
Early to bed and early to rise and you'll be groggy when everyone else is
wide awake.

Send instant messages to your online friends http://uk.messenger.yahoo.com 




More information about the Nottingham mailing list