I was thinking of creating a continued fraction class for us, but then I saw a note about how continued fraction comparison is easier than regular fraction comparison. More importantly, c.f. comparison is just list entry checking, no arithmetic operations. Conversion from r.f. to c.f. involves only division and remainders. These facts combine to make a r.f. comparison that doesn't need multiplications, and therefore avoids overflow.
My understanding is that continued fractions aren't very useful for anything that's involves actual arithmetic, but they can be used to solve specific problems like the one raised for rational. They're also extremely important for evaluating numeric approximations of various functions, but rather hard to compute unfortunately. Even so the modified Lenz's algorithm would be a useful addition to Boost at some point. John.