Haven’t there been successful rewrites of nontrivial applications? I’ve always been skeptical of Joel’s Netscape example. Was the rewrite really the mistake? Couldn’t the mistake have been making an application where no one knows how it’s supposed to behave?

I'll admit the example is a bit sketchy. My dim memory is that Netscape's mistake was that their business model was charging for their product, which is not unreasonable but was doomed when Microsoft gave Internet Explorer away for free and included it in Windows (so you'd really have to go out of your way to find Netscape's paid alternative).