Earman on underdetermination and empirical indistinguishability door