Actually, that raises an interesting point: Are there other libraries that could be used? That is, if the goal is diverse implementations, openssl and a fork of openssl are too little of a difference. There's bearssl which looks even more different (although then there's the question of being API-incompatible), and boringssl (which is... another openssl fork...). Or is this going to end up like browser engines where if you want to be compatible enough to swap in you have to implement 30 years of features which is to say that nobody will ever do it?