It's Raining Mainframes....

In Robert X. Cringely's thought-provoking book, Accidental Empires (Addison-Wesley, 1992,) he points out that the year 1999 will witness a mass die-off of older mainframes — not in terms of their being replaced by denser and quicker systems, which has been happening for years already, but because they abruptly lose their ability to process transactions. The reasoning behind this is various, but the simple case is that many older mainframes (especially if they’re running COBOL) store their dates in the format

YYMMDD

and, when YY returns as 00, will halt on error. "Hardly any programmer in 1959 expected his payroll application to be still cutting checks in 1999," says Cringely, "so nobody thought to teach many of these computer programs what to do when the calendar finally says it's the year 2000."

They didn't reckon with the principle of cybernetic inertia, which says that hardware in place tends to remain in place. Sure, there may be workarounds. But for lots of older computers, the programming overhead of dealing with this kink will be the last push over the cliff.

This hit the mainstream press in March 1995 when Guy Gugliotta, in the “Capitol Notebook” column of the Washington Post, summarized it with a deliciously dire spin. “This is no joke,” he opines; “....conversion to make federal computers read dates beyond Dec. 31, 1999, will cost an estimated $75 billion in the United States....In the end....you have to get deep into your computer system and change every line of code that uses a two-digit date field to one that has four digits. Better get started early.... [Conversion expert Chuck]" Ross estimates that the average federal agency will need to modify 75 to 100 applications, which should take 45,000 to 60,000 people-days.”

Cringely and Gugliotta are right; in the face of titanic investment for patches – to source that may not even still exist – mainframes will be scrapped wholesale, and the oldest first. From the standpoint of function it only makes sense, since the oldest hardware is usually the slowest. But to the historian and preservationist, the oldest hardware is often the most significant. If we intend to respond to this crisis, we have four years to make plans and marshal resources; four years to find and equip facilities; four years to nail down funding. And for a project of this size, four years actually means that we're in a screaming hurry.

What we're trying to do here can only be done once, or given up for lost. If you're reading this you can help, with a donation of whatever you can afford.

Save the mainframes!

Return to Introduction