On Kraichnan's 'direct interaction approximation' and Kolmogoroff's theory in two-dimensional plasma turbulence