Ted's objection is perfectly reasonable, and the way that Linux works and has always worked internally. If an internal (not userspace) API changes then code that uses that API has to be fixed. He also quite reasonably says that it'll take time to understand the trade-offs of encoding constraints into the type system.