I've added a small postscript explaining why calculus of variations (in its simplest form at least) is not applicable to this problem.