SAN FRANCISCO (03/03/2000) - To celebrate the release of Microsoft Corp.'s Windows 2000, today's theme is products that crash, fail and generally let us down. Of course Windows 2000 is supposed to have none of those flaws. If you can get it installed (set aside a full day), and if you can live with the side effects of it having been designed mainly for big corporate users (for instance, it won't let you dial in to America Online), it goes a long way toward reducing the crashes, failures and general annoyances of Windows 98.
Yet when Windows 98 came out, it was presented as the highly refined, much more stable replacement for the overly crash-prone Windows 95. And when Windows 95 made its debut, amid advertising and sloganeering worthy of a presidential campaign, it was depicted as the near-miraculous answer to the prayers of anyone unhappy with Windows 3.11. And when 3.11, and before it Windows 3.0, were released ...
That we are able to seem newly enthusiastic about products we know we'll soon consider subpar may indicate that software makers, including Microsoft, have mastered the "planned obsolescence" strategy of Detroit automakers in their prime. But it also reveals something deeper about the circumstances in which customers will and will not accept unreliable performance, which in turn has consequences for the Internet.
People are famously irrational in deciding what hazards to pay attention to.
More Americans were killed in bicycling accidents last year (about 900) than died in all U.S. airline crashes in the 1990s. But when a big plane goes down, there are big stories which we all read. Residents of Los Angeles or San Francisco are hundreds of times more likely to die in a household fall from a ladder than in an earthquake, yet the earthquake danger is what people are hundreds of times more likely to discuss. (In the 20th century, about 5,000 Americans died in earthquakes, most of them in San Francisco in 1906. That many people die in household falls every four months.)Our reaction to these different perils is both not quite logical and completely understandable, since we are affected by more than the pure statistical risk.
The other factors we inevitably weigh include the element of sudden horror quite high when the Earth opens up or you're trapped in a falling plane; an individual's ability to control the risk, which is close to zero for an airline passenger; and the ability to correct or recover from the damage after it has occurred, which is very low for airplane crashes.
A less dramatic version of this subjective risk assessment helps explain the market's tolerance for some kinds of high-tech failures and its acute sensitivity to others. Among modern inventions, by far the least reliable is the cell phone. I don't know of any official tabulation, but I would bet that at least one-quarter of all attempted cellular conversations end with a dropped call. Citizens would storm the Social Security Administration with pitchforks if it had an error rate one one-thousandth this high in mailing out checks, but we tolerate it with cell phones because the surprise factor is low and the ability to correct the damage is high. You just press Redial.
Computer operating systems are the next most unreliable modern product. While keeping count these last seven days, I have found that the two computers I run nearly around the clock a Dell desktop and an IBM laptop, both factory-equipped with Windows 98; have averaged a total of three hard freezes a day, requiring power-off resets. By comparison, my low-tech, totally unglamorous car; a 1989 Plymouth Acclaim; went not just seven days but more than 10 years with no mechanical defects. If the car were constantly breaking, we'd certainly blame stodgy, "old-economy" Detroit. But we tolerate failure in more "advanced" products partly because we have no alternative, and also because of mitigation tools now built into most software. The two very best features Microsoft now offers are Control-Z, or Undo, and Auto Recovery for files you were working on just before a crash.
So what about the Net? One aspect of its operation still exists in a fault-tolerant zone like that for Windows and cellular phones. Customers know that ISPs and servers may at any moment brown out, slow down or go offline.
Customers accept it because (for now) they have to, but also because they expect the recovery to be fairly prompt.
But in two other areas, a pattern of failure could really harm the Internet Economy. The first is the handling of credit cards. Only four years ago, the accepted view was that people would never feel safe sending such intimate data over computer networks. (The same thing happens each time they charge a purchase at a store, but never mind.) With amazing speed people have overcome that discomfort, but the fears would return just as quickly with the first significant reports of card numbers being stolen or misused. This is the online equivalent to a potential air disaster, because the public's threshold for acceptable risk is "illogically" low. The odds of dying on a commercial airplane are well under one in a million, and some people still worry about it.
The Net's handling of credit cards needs to remain about that safe or consumers will shun e-commerce as if it were Valu-Jet.
The other danger involves the low-tech Achilles' heel of the Net: delivery and real-world fulfillment of the deals carried out electronically. When I haven't been counting how often my computers crash, I've been making a list of things I order online that don't arrive on time. I have learned what will come through as scheduled. Four orders from Drugstore.com all arrived when promised. For computer hardware from CDW.com, two for two came on schedule. United Airlines has not screwed up any of the tickets I've bought on its site.
But those are the exceptions. Travelscape.com made a hotel booking and charged it to my credit card but neglected to say anything about it to the hotel, as I discovered when I dragged into the lobby on a rainy night. I planned an e-commerce Christmas, with boxes of fruit, smoked salmon and regional souvenirs ordered in November from several dot-coms for far-flung relatives. None of the presents arrived in December; some showed up in January; the rest were shipped in February. Game equipment from specialty dealer 4CDS.com arrived a month late and missing a necessary part. I made upgrade requests and flight queries via Northwest Airlines' Web site, but I figured that it must be a dummy page or a hoax, since there was never a response of any sort. I ordered an elaborate electronic postage scale from Staples.com as a Valentine's Day present for my wife. (Yes, they do call me "the Romantic.") Delivery was scheduled for Jan.
25. I called a week after that to see where it was. A week after that a nice box arrived, which, when my wife opened it, proved to contain not a scale but three steno pads, apparent value $4.50. And this just scratches the surface.
Most of the business I've done online this year has made me wish I hadn't.
Here the risk calculation is different from that of airplanes or credit cards.
The problem is not that some infinitesimal hazard might be exaggerated. It is that reasonable people will decide that the delay and error are not worth it, so instead they'll return to their old-world customs of driving to the mall or ordering through an 800 number.
For now, I'm still an optimist. I've just ordered the Plymouth Acclaim's successor, online. Stay tuned for the results.
James Fallows writes about technology and politics from Seattle.