The affine separation problem revisited