In this paper, I want to propose a revised version to supervaluationists' strategy on vagueness by rough set theory. The main reason to my work is due to the very intuition about all the boundaries of vague terms must be obviously vague, i.e. not only the boundary of positive and negative extension but also the boundaries of positive or negative extension and borderline cases are vague. Nonetheless, although supervaluationists indeed provided a smart way to solve the first-order vagueness but they were trapped on another serious problem, i.e. higher-order vagueness. I hold that the rough set theory can save supervaluationism from the big trouble of higher-order vagueness. The central idea of preceding claim is that the boundaries constructed by rough set theory are not fixed but rather flexible, which fit to the essential features of vague terms.