A plain man's view of the muffin-tin approximation in molecular calculations, and the solid state