On Besicovitch's 1/2 problem