Scaling of a set of values that are almost, but not quite the same

[optseb]

Imagine 3 double precision scalars, one of which is greater than the other 2 by std::numeric_limits::epsilon(); The output of this when auto scaled by morph::Scale would be { 0, 0, 1 }. Does it make sense to consider all of these more or less then same (perhaps by computing the variance)? Would the desired output be { 1, 1, 1 }?