Auld Lang Syne

An article about the Voyager Deep Space Probes circulated recently on Facebook; after 40 years of flight, management was looking for programmers who knew Fortran, COBOL, and assembly language. The call for programmers proficient in these "ancient" languages garnered a lot of laughs from the younger programmers in our community but demonstrates a future problem that is just beginning to appear.

The one remaining original programmer on the Voyager project, Larry Zottarelli, is now 80 years old. Remember that Voyager has been flying for about 40 years, and its design really began perhaps five years before it was manufactured, tested, and launched, when Zottarelli was still fairly young. Probably a lot of the people working on Voyager did not think it would still be working 38 years later and would not believe that it might still be operating well into the 2020s. NASA needs programmers who can keep the software "alive."

This is not the first time I have run into situations of ancient hardware and software, but it is one of the most interesting because the hardware cannot be upgraded.

Many times, the solution is simply to develop a new system, run it in parallel with the old system to make sure they create the same answers, then discard the old system. This process can generate maintenance cost savings that pay for the redesign and re-implementation of the new system.

Another NASA project was the redesign of the Johnson Space Flight control center more than 20 years ago. The previous control center had been built more than 20 years before THAT, and its hardware was hopelessly out of date. Spare parts were hard to find, and each year the cost of maintenance went up. At the time I became involved, NASA was paying more than $200 million a year just on the maintenance of their ancient hardware and software.

A project was proposed to rewrite all the software and replace the hardware with a "standard operating system" and modern production hardware. The cost of this replacement, which gave much better reliability and flexibility than the old systems, would be (coincidentally) $200 million, the same as one year of maintenance on the old systems. However, the estimated yearly maintenance cost would "only" be $23 million and, more importantly, could be accomplished by younger engineers using more modern tools.

The use of these younger engineers was not (as some people commenting on the Voyager article thought) to save money by hiring younger, lower paid, instead of older, higher paid engineers, but to have the luxury of using engineers who were not on the verge of retiring or (worse yet) dying. It is one thing to bring back an engineer from retirement to fix something and quite another to try and bring them back from the dead.

The new system was designed and built using the Digital Unix operating system and DEC Alpha systems hardware. Two hundred million hours of telemetry data was fed through both the old and new systems to make sure the new system worked exactly the same as the old system.

A third "oldie" story has to do with automated teller machines (i.e., ATMs) used by banks. Many of these machines were designed in the days of the Intel x386 processors and had been functioning very well for a decade or more using IBM's OS/2 operating system. Then IBM dropped support for OS/2, and the ATMs were left without anyone to provide software maintenance.

At the time this happened, Microsoft had also dropped support for the x386 computers, moving on to x486 and Pentium systems. Even if Microsoft had still supported the base hardware, the ATM machines typically did not have much memory in them, so modern Microsoft Windows systems did not work either. A lot of these banks chose GNU/Linux, not only because it did the job, but because they knew they would always be able to fix the problems that occurred.

Of course, there was also the unforgettable campaigns around the "Y2K problem" [1], which has probably already been forgotten by some of the younger IT crowd, along with the corresponding "End of the Unix Epoch" [2] that will be happening in a few years.

The problems we have not encountered yet will be the most interesting. We are dealing with a relatively young industry by historical standards, and some of the aging issues are just beginning to show up.

Do not, however, laugh at the "ancient languages," my young friends. Someday your favorite language may also be old, deprecated, and ancient.

Except for C and GNU/Linux, of course.