I'm generally a big fan of Joel Spolsky, but in retrospect, I think this advice is just wrong, and I think Skype is a perfect counterpoint. That is, sometimes a rewrite is a horrible idea, but at the same time sometimes not doing a rewrite is a horrible idea. If making changes to the code becomes such a nightmare that your rate of progress is much less than your competitors, you're going to lose.

While there is still some good advice in that blog post, hard-and-fast rules are rarely correct. Most things in engineering are tradeoffs, and it's tough to know sometimes what the right balance is.