Tim> How many trailing zeroes do you expect to save this way? The real point is to have small, efficient rational representations for floating-point numbers, even those with large exponents.