This paper gives a method for constructing a grid of points from scattered data; in the examples, though, the author assumes that the data points are a subset of the grid points. The grid values minimize a certain discrete pseudonorm; this process may involve approximations to derivatives. The method does not construct an underlying surface, and the approximation thus depends on the chosen grid. The examples seem reasonable, but several points worried me. Because no boundary conditions are applied, it appears that the formulation of the equations often leads to a singular (but consistent) system of equations. The solution obtained seems to depend on the initial guess and the solution method. When faults are imposed, the scheme may leave certain grid values undefined. These issues must be resolved before the method can be seriously applied.