Order of error for rational approximation of irrationals
Widening the question a little... is there any reason why continued fractions cannot be used? They'd give the best rational approximation for the smallest denominator, and the error bounds are well defined.
