Key points are not available for this paper at this time.
We show how first order optimality conditions for a very general nonlinear optimization problem may be derived in a conceptually simple and unified manner in terms of certain multivalued functions associated with the problem. Necessary conditions for general problems and sufficient conditions for convex problems are developed, and the classical multiplier conditions are shown to be related in a simple way to these.
Stephen M. Robinson (Tue,) studied this question.