This award adds an experimental component to ongoing research on exploiting type systems as a basis for interfacing dissimilar programming languages. The experimentation will apply fundamental computer science insights to support software for parallel numeric supercomputing. Three specific new efforts are included: 1. Adaption of polymorphic type checking and inference, as found in Standard ML, to deal with vectors, arrays and various numeric data types as in Fortran 77; 2. Evaluation of "mixin" based inheritance as a framework for structuring C-based numeric libraries into more modern and secure C++ clauses and 3. Exploitaion of declarative language concepts as a high level "harness" for parallelizing large blocks of existing imperative code. This work will be performed while the Principal Investigator is on sabbatical leave at Cornell University. this opportunity will take advantage of several unique resources: (i) collaboration with research groups in constructive type theory, language design for physical and mathematical scientists, and applications of dataflow programming; (ii) interaction with the staff and users of the Cornell Center for Theory and Simulation in Science and Engineering, and (iii) development of prototype software on Cornell's 48-node BBN Butterfly multiprocessor.