A method of encoding binary floating-point numbers and decimal floating-point numbers within the limits of finite precision available on computers.