The title problem is considered from the viewpoint of one involved in aerodynamics problems, rather than reactor modeling. The status of opinion is briefly reviewed on two sources of error: errors due to discretization, especially in regard to “high cell Reynolds number” difficulties, and errors in the continuum equations, especially in regard to turbulence modeling. It is concluded that significant improvement is to be expected over the lumped-parameter codes, but that even the forthcoming codes based on simulation of the partial differential equations are expected to be limited to a rather crude level of accuracy.