This paper was discussed on HN three years ago (
https://news.ycombinator.com/item?id=2379098), but I'm resubmitting because it's been a while, I figure most people haven't seen it, and it's a great paper. It really got me thinking when I first read it. Specifically on the notion the issue of: how sure are you, really, that the compiler you implement is correct? As someone who works on a programming language, the question is not merely academic.