Just read the introduction, and it hits all the right spots in explaining why people (including me) have failed to come up with good subtyping in Hindley-Milner

Just read the introduction, and it hits all the right spots in explaining why people (including me) have failed to come up with good subtyping in Hindley-Milner world. I expect people to rush to update their various HM-based systems to this new approach. This seems like a really big thing to me.