There is no need to show the importance of arc-consistency in Constraint Networks. Mohr and Henderson [9] have proposed AC-4, an algorithm with optimal worst-case time complexity. But it has two drawbacks: its space complexity and average time complexity. In problems with many solutions, where constraints are large, these drawbacks become so important that users often replace AC-4 by AC-3 [8], a non-optimal algorithm. In this paper, we propose a new algorithm, AC-6, which keeps the optimal worst-case time complexity of AC-4 while working out the drawback of space complexity. Moreover, the average time complexity of AC-6 is optimal for constraint networks where nothing is known about the constraint semantics.