The impact of parallelism on software

There seems to be a tug-of-war raging between computer procurement technical evaluation committees, computer designers and scholars of computer science and numerical analysis over the issue of parallelism in computations. Caught in the middle seems to be the user community and the compiler writers. For the scope of this paper, the term "user community" will be assumed to be Fortran programmers who are involved in solving problems that require large computer resources, e.g. plasma research, weather prediction, ray tracing, seismic analysis, econometric modeling, weapons research, reactor calculation, etc.