There is a widespread opinion that ice surfaces are covered by a liquid layer at equilibrium, and several treatments of the ice surface describe it as a homogeneous layer with a definable thickness. Arguments are presented that the ice surface cannot have a homogeneous surface layer and that representations of it that use such a layer are not useful approximations. No part of the transition layer on ice can be homogeneous except at a triple point without violating the phase rule. Some experiments suggest the existence of a surprisingly thick transition region at ice surfaces, but others do not. Water apparently does not completely wet ice, since its contact angle on ice close to the triple point appears not to be zero, and sharply faceted growth forms of ice from the vapor close to the melting point also suggest the absence of a thick, liquid, surface transition region.