Re: 4 the FAQ: Are Commercial DBMS Truly Relational?

From: Kenneth Downs <firstinit.lastname_at_lastnameplusfam.net>
Date: Sun, 10 Oct 2004 13:21:31 -0400
Message-ID: <r2rbkc.5b8.ln_at_mercury.downsfam.net>


Laconic2 wrote:

>
> There is a widespread belief in this field that wrong answers with good
> performance are closer to the goal line than correct answers with poor
> performance. I will never come around to that point of view. I almost
> always want to get it right, first, then work on getting it right, and
> fast.

Have you ever "optimized" the work of others after the fact? I was in a situation once where I optimized several programs, gaining performance improvements in every case counting between 1-3 orders of magnitude. The supervisor was astonished and asked me how I did it. In each case the answer was the same, the program was doing too much work. It would do a lot of wrong work, and then somewhere do the right work. The entire optimization effort was in reducing the program to doing only what was necessary. There may have been a couple of tricks-of-the-trade for the platform in question, but mostly it was eliminating work.

This lead me to the hypothesis that the correct answer is as fast as you can get, simply because you are doing exactly the work you are supposed to do, and nothing extra. My group adopted the motto that we wanted to go 'as fast as the disk', anything faster was impossible, and anything slower meant we were making some horrible mistake, since RAM and CPU are so much faster than disk.

Next it was my turn to be astonished. I suggested to the supervisor that more effort go into the beginning of these programs, to ensure that they were correct, and that would eliminate a huge optimization effort afterwards. He did not agree.

-- 
Kenneth Downs
Use first initial plus last name at last name plus literal "fam.net" to
email me
Received on Sun Oct 10 2004 - 19:21:31 CEST

Original text of this message