Underdetermination, Realism and Empirical Equivalence