“…However, the methods used in [13] and [12] cannot tell us whether we actually have c p < 1, that is, whether the maximal operator M has a smoothing effect on f . For p = 1, Theorem 2.5 of [2] states that DMf 1 Df 1 , and c 1 = 1 is sharp, while for p = ∞, we have DMf ∞ ( √ 2 − 1) Df ∞ and c ∞ = ( √ 2 − 1) is best possible, by [1]. Thus, it is natural to conjecture "by interpolation" that whenever 1 < p < ∞, the optimal constant c p satisfies c p < 1, and furthermore, lim p→∞ c p = √ 2 − 1.…”