Software

On the software front, the news is mixed. A bad trend is that we are finding more software packages to be helpful and even essential, and thus we have more components overall. A good trend is that vendors are working to increase compiler and operating system compatibility. And we are gaining more experience at building platform independent software. SEP has 4 binary platforms. The book is fully operational on the Sun-4 and the Convex. I have not addressed the byte-swapping issue required for complete compatibility with our DEC-3100. The books would probably run on our many Sun-3's but they are old and slow and some parts of their environment are degraded and not worth maintaining, except for book display.

cake.

Dave Nichols reports that cake worked immediately on all our platforms.

ratfor

is provided by AT&T, the inventor of UNIX. It was once part of the standard package, but has become an added cost extra, and it is not supported at all by some vendors, such as Convex. AT&T will sell the source, but not cheaply, and last I heard, they give no support. On the other hand, the DEC Ultrix fortran compiler will compile from either fortran or ratfor source.

seplib

is the most popular I/O library invented and used at SEP. It is also formally supported by a group of French sponsoring companies, and is informally supported by SEP alumni at various sponsoring companies. Seplib is deeply dependent on internals of the UNIX operating system, and thanks to Stew Levin and Dave Nichols, it has survived the arrival of RISC, POSIX, and ANSI-C. Fortunately, we are getting organized with Benoit de Boullay at the French Petroleum Institute to try to maintain it in an orderly way across more hardware platforms and prevent future environmental traumas. (Seplib is best documented by its on-line manual pages which might appear in this report. About the only description of it in SEP reports is found peripherally in Claerbout, 1986.)

vplot

is SEP's intermediate plot language which all our programs write, which can be interpreted on all our devices. Vplot is principally the work of Joe Dellinger, though it is based on early work of Rob Clayton and Dave Hale. It seems to be robust and stable, though there are many opportunities for disruptions with future operating systems. The biggest current problem is that the filter from vplot to Xwindows has some aggravating bugs. Rick Ottolini wrote one version called X11pen, but the most popular version here now is Xvpen, written by Steve Cole in XView.

TEX and L^ATEX

are fine word processing software, free I think, and they are getting easier to install. But the fonts can take a lot of space.

XView

is a freely distributable tool kit from Sun Microsystems that is needed for the interactive programs.

X Windows

is the key piece of software that has made our multi-platform environment possible. But it is not problem-free. Installation and maintenance are not trivial, especially when trying to support the various vendor-specific versions of X (DECwindows, Sun OpenWindows) as well as the portable MIT version.

I hope the above list of troublesome maintenance areas is complete, but fear there will always be another snag.