Joel Spolsky is becoming an old fart

On his latest article, Joel makes so many mistakes that I kinda felt sorry for him. On the article he talks about his experience on optimizing compiler speed at Fog Creek by upgrading the computer hard drive.

First, against all common knowledge about compiler speed optimization, he decided to improve disk speed. Most - if not all - current solutions on compiler optimization involve some sort of distributed CPU effort (distcc like). I am all for fighting common knowledge, but you must be able to out-smart everyone, otherwise it's just depressing. He tackles the optimization problem without profiling what the bottleneck is, a beginner's mistake.

Then, instead of using a standard machine, he used an old laptop. After spending days trying to copy a HD to a SSD, he realizes the new SSD improves the computer speed (!), but that compiling time is still the same. More than that, he concludes that CPU is a bottleneck, which is the correct conclusion but one that should not have been concluded from his experiment, since he used an old machine that has obvious less CPU available.

Summing up: he didn't think enough before acting and because of that he did a bunch of useless work, obtained the wrong conclusion from his failure and is proud enough of that to write an article on it. An outstanding manager.