There have been a number of other proposed standards for input. The most popular approaches use a high-level modelling language to specify problems. Typical examples are GAMS [1], AMPL [7] and OMP [4]. Such approaches are useful for specifying repetitious structures, but do not really attempt to cope with useful nonlinear structure (like invariant subspaces). Recent work [8] hopes to overcome this disadvantage.
We have recently become aware of other suggestions for the input of large-scale structured problems. These proposals are based upon representing nonlinear functions in their factorable [13] or functional forms [15]. Such forms are the the logical extensions of (2.1) in which a function is decomposed completely into basic building blocks. The advantage of such schemes is the potential for the automatic calculation of derivatives, but this must be weighed against the difficulty of describing how the building blocks are assembled. We await further details of these interesting proposals.
