SEP is investing considerable effort in SEPlib90 in order to be able to do effective research into 3-D seismic problems. We think it is a good design and are committed to it. One of the early discoveries made as SEPlib90 was born was that ``overhead'' coding had grown enormously. A great deal more programming, much of it predictable from program to program, was required to deal with header axes, grid axes, and the like. Fortran's lack of user-defined structures and operators engendered this as well as the need for declaring, passing, and keeping track of many new variables to deal with headers and grids. SEP knew fairly early that some other language was necessary, and we in particular chose Fortran90 for its familiarity, versatility, and its promised power - we had just taken delivery of a 16 node SGI power challenge, a platform where much attention has been paid to Fortran90 and its use as a parallel language.
The framework is in place now for use of Fortran90 at SEPlib90, though there remain numerous kinks. For the future we plan further development; a full set of operators, more routines for easily constructing the spaces needed by filters and residuals as necessary for inverse problems, and whatever else our experiences tell us we need.