Y2K Problems will be minor

Predictions of apocolypse fill the daily news, uncritically accepted and then abandoned without responsibility. The world will run out of resources by 1900. By 1950. By 2000. The war against Al Queda in Afganistan will be "another Viet Nam." The war against pan-arabicist Saddam Hussein in Iraq will be an "endless morass." And, lest we forget, the world will end on January 1, 2000. Oh, that's three years ago. So, I reprint here my predictions of 1999:

I'm not expecting a Year 2000 cataclysm. I've read a fair amount of literature on the subject, and it largely breaks down into two categories. In the first, you find sober discussions of the details with a conclusion that the problem is largely manageable. Any well-run company has already addressed the problem and is well on its way to minimizing its effects. But then there is the second category, in which you find Armageddonites whose argument is essentially that all computers operate by hidden processes and are interconnected in unknowable ways, so we cannot possibly predict which computers will fail and a small failure by a seemingly minor device will lead to a chain reaction destroying the world.

This second category gives metaphysical power to the unknown and denies obvious human ability to plan ahead.

We have substantial evidence against the "interconnected-chain-reaction" theory. Nearly all significant programs that have ever been written contain errors of various sorts and occasionally fail. These typically lead to very localized consequences. Occasionally the failure affects several connected systems. Only very rarely does an error influence a large number of computers; and only when either all the affected computers were running the same program, or the failure was deliberately caused by human malice. The Y2K sort of programming error does not differ from other programming errors in any way that makes it more likely to contaminate other computers. Obviously Y2K errors will be concentrated in a narrow period of time, but this is the effect of many individual errors all happening contemporaneously, not the result of Y2K errors having an unusual ability to spread.

Let me make an analogy: Suppose we discovered a flaw in the little computers used in cars such that an unusual number of cars would develop engine trouble in January of the year 2000. There would obviously be costs and inconveniences resulting from this; business at gas stations would suffer; people would more often be late for work; some businesses would be hurt if critical deliveries were missed; there might even be a few related fatalities. None of these things are good. However, it would be insane to think that the problem is contagious, that filling your car up at a gas station could infect the pumps, which would somehow spread the disease back to the main computers of the oil company, and from there to its banks and suppliers, where it would reach out to anyone who telephoned the banks, and on to every institution that uses computers until the entire civilized world was thrown into a state of paralysis and chaos. Yet, this is the claim of the chain-reaction allegation.

Y2K errors differ from other errors in only two ways: They are more limited in time and we can describe their cause and cure much more easily and precisely. By analogy to weather, the Y2K errors are not like knowing that there will be some unpleasant weather someday, they are like knowing that there is a 30% chance of rain next Saturday. Against which can you prepare better?

If a company is badly hurt by Y2K problems, it would be because that company's management has not addressed the problem despite the decades of advertising. (This is, after all, an event whose coming is not unexpected.) The mismanagement of the company will not have been limited to this isolated issue. This will already be a company that you don't want to deal with as an employee, customer or investor.

Of course, good companies sometimes depend on less-good suppliers. But they will be prepared for missed deliveries and billing errors, maybe even a bankruptcy or two. The good companies will be scarcely affected, no more than by some unusually bad weather.

But what about government monopolies, such as cable TV, water or electricity? I don't know about you, but my cable TV goes out for days at a time, and I'm still alive. Even my electricity fails. Last year I had ten days with no electricity and too much snow to even drive out of my house. I'm still alive. Could the electric systems and water systems fail due to mistakes in programming? I suspect they already do, and the consequences are localized and relatively minor.

So I talked to my local water commissioner. He showed me the small water plant we use, and it doesn't even have a computer in it. Computers are not reliable enough, he explained. They suffer from all sorts of delicate troubles. So the plant uses large, old-fashioned relay boards, the kind that have kept factories operating for a century.

What about government itself? Police and fire departments dispatch systems will largely continue to work, though in a few badly-managed cities they may fail. You probably don't live there. If you do, deal with it. Airlines likely will be shut down for a while because no government official will want to be the one who certified the air traffic computers as safe. So don't plan any important trips near the end of this year ("99" is also a risky number) or next. Try not to be out of the country. Do your Christmas shopping a little early. And maybe the Social Security office or the IRS computers will fail. Is that bad?

What about the "embedded systems" often mentioned (small computers built into other equipment)? For example, my car has several little computers in it, so does my VCR, television, thermostat, light timers, washing machine, dishwasher, lawn watering timer, and even the electric generator I bought after last winter's storms. Are these vulnerable to Y2K problems? In the main: No.

Most of these computers don't rely on the date in any way. Of those that do, for example the VCR, they primarily need it to compute the right number of days in February. However, the year 2000 is a leap year (an exception to the normal rule that years divisible by 100 are not leap years). So sloppy programming would get lucky and the right answer. But even if my VCR missed a recording, or my lawn got watered off-schedule, or all the lights in my office building turned off at the wrong time, I wouldn't die.

Some embedded systems used in manufacturing or military applications pose a larger problem. These are often programmed to stop working after a certain number of years, signalling that it is time to replace them with newer devices. Many of these have not been tested for the effect of Y2K dates, and very likely some will fail. Some failures will be costly, manufacturing lines will shut down or destroy product.

Another likely cause of failures will be in the interchange of files between programs that use more space for dates and older programs not prepared for this. Some older programs will make erroneous calculations, and it is hard to test all of these older programs in advance.

So, certainly there will be problems. But they will be limited, non-contagious, and largely eliminated from the most critical areas (excepting government) and the best companies.

For most of us, with a little preparation, this isn't going to be any worse than a bad winter storm.

In retrospect, the problems were even more minor than I had predicted. To those who do not accept faith, no explanation is needed. To those of faith, no explanation is possible.

Chicago and Northwestern Railway, map of Chicago area, 1904

Chicago and Northwestern Railway, map of Chicago area, 1904