How do you pick the 64 bits of redundancy for RNR, is it coming from the field size/implementation details, or is that just the point where the leakage tests stopped looking scary?