General purpose nonlinear programming package. This paper tries to make three steps towards more efficient applicability of mathematical programming algorithms : First, a program package is presented which easily allows comparisons of different algorithms applied to a specific minimization problem. All the algorithms use similar, problem-dependent termination criteria. Secondly, an efficient standard algorithm for unconstrained optimization without using derivatives is described. It is a quasi-Newton method in which the termination criterion of the unidimensional searches becomes refined according to the convergence of the multidimensional optimization. The required gradients are evaluated by numerical differences using step-lengths which iteratively are optimized to get maximum accuracy. Thirdly, the user of the nonlinear programming package has the option to leave the selection of the optimization method up to the program. Numerical examples are presented.