Yeah I honestly don't know what I would change. Maybe replace some of the control characters with more common characters to save a tiny bit of space, if we were to go completely wild and break Unicode backward compatibility too. As a generic multi byte character encoding format, it seems completely optimal even in isolation.