Abstract: Suppose that there are two pass-fail components whose reliabilities are p₁, p₂ respectively. Let us assume that the test numbers are N₁, N₂ and the success numbers are S₁, S₂. It is the goal of this article to discuss the classical exact lower confidence limit of difference θ=p₁-p₂.In this paper we give the ordering rule of sample points and develop a method finding classical exact lower confidence limit of θ by using the ordering method of sample points. Morever we prove the exactness and optimality of our results.