Let D be a domain in C n with smooth (that is, C ∞ ) boundary. If 0 ≤ α ≤ ∞ we denote by A α (D) the space of functions holomorphic on D and of class C α on D. We write A(D) for A 0 (D) and A ω (D) for the space of functions holomorphic on a neighborhood of D. We say that a point p ∈ ∂D is a peak point relative to D for A α (D) if there exists a function f ∈ A α (D) so that f (p) = 1 and |f | < 1 on D \ {p}. We call f a peak function. This condition is clearly equivalent to the existence of a strong support function, that is, a function g ∈ A α (D) so that g(p) = 0 and Re g > 0 on D \ {p}. We say that p ∈ ∂D is a local peak point forWe want to determine whether boundary points are peak points. Finding a peak function can be thought of as giving a quantitative converse to the maximum modulus principle. Now observe that if f is a peak function at p relative to D then 1/(1 − f ) is a holomorphic function on D with no holomorphic extension past p. Thus if every boundary point of D is a peak point then D is a domain of holomorphy. In light of this fact, if we want to determine whether boundary points are peak points it makes sense to restrict our attention to domains of holomorphy. By the solution of the Levi problem these are the (Levi) pseudoconvex domains. Here is another observation: If there is a complex disc in the boundary of D then (by the maximum modulus principle) no point in the relative interior of that disc can be a peak point for A(D). A natural condition in this context is that D be of finite type at the point p in question, in the sense of D'Angelo: There is a finite upper bound on the order of contact of nontrivial complex analytic varieties with the boundary at p. The infimum of all such upper bounds is called the type at p. We formulate the major open question in terms of this notion.