In this report, the authors have exposed an economically important but difficult problem: the salvaging of subprogram libraries written using older programming languages so that they are usable with recently developed programming languages. They succinctly describe just how nasty the problem is by classifying the difficulties into two categories: those due to differences in language definitions, and those arising from incompatible implementations.
They give some examples of “mixed language programming” between FORTRAN and VAX/VMS PASCAL, FORTRAN and DEC System 10 SIMULA, FORTRAN and Control Data ALGOL 68, and FORTRAN and ADA. This should not lead one to believe that the root of the difficulty is FORTRAN, but that subprogram libraries are usually implemented in that language.
In the last section, they suggest approaches to automatically facilitate mixed language programming based on ideas of modular structure and the preservation of interface information between compilations. There are not many solutions here, just some food for thought. Their suggestions for automatic translation of source code may itself not be economically feasible. An acquaintance estimates that, in order to automatically “lift” the FORTRAN 77 codes at his installation into even a modest subset of the language currently called FORTRAN 8X, 20 man-years of effort must be expended] This paper is most useful in delineating the problem, but will not be of interest to readers hoping for an easy solution.