I agree - type overloading says templates to me rather than inheritance. C++'s tuple class lets you do some of the cool listy type things you can do in Haskell. Using typedef tuple<double, double> FloatPoint and taking it from there seems a lot more like the Haskell example and would be less code - one line each for the typedefs and a one-line function to implement <<. (Not taking anything away from Haskell, which I'm completely in love with.)
The typedef tuple strategy doesn't let you distinguish between types except by the types and ordering of their fields. If you later have a constructor for another ADT (or even the same one) which also takes two Float arguments, you end up with multiple typedefs for tuple<double, double>.
