> Fractional positions should be represented using arbitrary-precision decimals so that they don't run out of precision. Floating-point numbers are insufficient.