> I really have to question the wisdom of adding implementations of > Curve25519 and/or Ed25519 that are not completely constant- time. It's unlikely to happen.