The authors consider the problem of fitting a linear combination of functions to observed data in a multivariate model. They discuss combinations of approximations in the l1 and l2 norm. A Fortran program that carries out the calculations is presented. An illustrative numerical example is reported, and different approaches are compared for this example. The resulting errors were very similar, but the corresponding regression coefficients were different to a certain degree, depending on the method used. This paper should be accessible for a wide audience. It is well organized and well written. The list of references is adequate and contains 18 items.