To: ayahuasca who wrote (18047 ) 6/5/1998 2:45:00 PM From: S.C. Barnard Respond to of 31646
Friday June 5 9:45 AM EDT Experts look to 1999 for preview of millennium bug By Richard Melville NEW YORK (Reuters) - In what looms as perhaps the greatest global finger-crossing exercise of the computer age, the summer of 1999 will mark the start of the end-game in the countdown to the new millennium. By then, as is now widely known, little time will remain before companies worldwide will know if they have succeeded in dodging a year 2000 bug that threatens to paralyze systems. What fewer know is that many experts believe the first victims will already have succumbed. "In fact, January 1, 1999 is likely to be a pretty good indicator of what to expect, because so many business models look forward one year," said Jim Woodward, senior vice president at Cap Gemini America, the U.S. unit of the international information technology services firm. Also, while the popular image of the year 2000 computer problem has centred around fear of a January 1, 2000 global systems crash, the reality is that a variety of short-sighted coding decisions adopted decades ago are coming home to roost before that date. The presence of these programming wrinkles -- many of which were born out of innocent habits -- means millennium compliance will be far more than teaching a computer that a year ending in two zeroes may follow one ending in two nines. One such potential problem lies in an old programmers' habit of using certain hard-coded field values to trigger specific functions or designate a file or routine end. Among the most popular: 9999, a number that corresponds to September 9, 1999, a date that must have seemed unimaginable given the 1970s vintage of some of the Cobol-language code that still survives on many mainframes. Another obstacle appears just two months into 2000, because through an ironic twist of timing, a rarely invoked leap year clause in the Julian calendar rules gets invoked. An exercise in approximation, the Julian calendar needs an extra day every four years to offset a slight imbalance between the length of a year and the amount of time in an earthly revolution around the sun. But that adjustment overshoots the mark, so every century, on a '00 year, leap year is cancelled. But, of course, even that adjustment needs an adjustment every 400 years, which brings us back to the year 2000, which becomes the first double-nought leap year since 1600. Got it? Don't worry, neither did a lot of programmers, apparently. "I've sat around a table with programmers and argued until we're blue in the face about what the rules are," said Jeff Richards, director of international sales and marketing for ConSyGen Inc, a Phoenix, Arizona-based code correction service provider. The possible result is that, once again, some code will revert to a 1900-style set of possible dates for the year and fail to understand the need for a February 29. This bug has already reportedly affected some personal computer software, making it one of the first year 2000 bugs to actually surface. There are plenty more problems to choose from. Remediation efforts may give rise to incompatibilities. One popular technique, called "windowing," involves designating a two-digit year value, 30 for example, as a cut-off point. Anything below 30 and the year is assumed to occur after 2000, anything higher and a 19- prefix is assumed. Of course, systems often include a multitude of individual programs that could use different window values, meaning a whole new family of irregularities are waiting to emerge. Also lurking are euro compliance, the Dow Jones Industrial Average's eventual ascent beyond 10,000 -- reportedly also an issue for some computers that only allow for a four-digit Dow-and a host of other bugs that should keep companies busy for some time. "We, like everybody else believe the time limit on disasters will be longer than January 1, 2000," said Richards. (Reuters/Wired)