FAA Grapples with Year 2000 Snafu

While the FAA's 23 million lines of code are being made Y2K compliant, the problem is also evident in the microcode that is built into some of its hardware.

As the year 2000 clock ticks away, computers around the globe are being tested for potential data meltdowns and other snafus. But while the clock's advance may be just another computer glitch for some organizations, it could present an outright disaster for others.

The Y2K problem - some computer systems record the year using only the last two digits, and thus at 2000 will malfunction in some way - recently claimed a high-profile casualty in the FAA's air traffic control system.

With an estimated US$91 million cost for the entire project, the Federal Aviation Administration has been at work on the problem for some time. But the FAA is now scrambling to find and eliminate these Year 2000, or Y2K, problems from aging - yet mission-critical - air traffic control computers. Even if that means completely replacing them.

"We're on one hand working to get those computers Year 2000 compliant, but at the same time we're working on replacing those computers," said Paul Takemoto, a spokesman for the FAA.

Takemoto said that the FAA is currently in the assessment phase of the project, where it will be determined exactly which systems and which lines of code need to be fixed.

"The host computer is the computer that's used at the 20 Air Route Traffic Control Centers," said Takemoto. While take-offs and landings are handled by an airport's air traffic control tower and another airport control handling the airspace within roughly a 50-mile radius of the airport, these Air Route Traffic Control Centers - or ARTCCs - handle all flights everywhere else.

There are 20 such ARTCCs throughout the country, each with a host computer and a backup system. All 40 of these machines - mid-'80s vintage IBM 3083 mainframes - are affected. The decision on whether or not to replace some of these computers, Takemoto said, will be made in 90 to 120 days.

But here's the catch - it is not entirely a software problem. While the FAA's 23 million lines of code are being made Y2K compliant, the problem is also evident in the microcode that is built into some of its equipment's hardware.

Microcode - or firmware, as it is sometimes called - are the modifiable programming instructions written in a processor's machine language that are part of the device itself. This makes the situation difficult.

"The product has not been sold since 1988 and is being withdrawn from service effective September 1998," said a technician at the IBM Year 2000 Technical Support Center. The technician provided an IBM Y2K product readiness report for the 3083 series, which showed that none of this equipment was Y2K ready, nor would it ever be - since it was being withdrawn from service.

Takemoto said that the FAA is going to attempt to fix the machines itself.

"They're talking about trying to fix it, just in case they can't get anything else running between now and then," said Peter G. Neumann, principal scientist at the SRI International Computer Science Laboratory, and moderator of the well-known RISKS Forum newsgroup. "They're in serious trouble," he said

Neumann said that there have been proposals to ground all aircraft for at least the 24-hour period beginning when the first time zone in the world ticks over to 2000 - and possibly shutting down the air traffic control system for even longer than that.

"Even after 1 January," said Neumann, "if you start flying and this bug is residual, you can have problems. The strategy would be that you'd do very controlled, careful experiments after the date hits before you enable flights to return to normal."

Neumann, whose RISKS Forum details potential dangers that computers pose to humans, believes that the Y2K bug is a potentially serious problem.

"We'd thought about it in 1965 when we were doing Multics," said Neumann, who was one of its developers. (Multics is an early - yet still in use - operating system.)

"Lots of folks have been talking about it - working on it - for a long time," Neumann said. "In Multics, we decided we would do something that would avoid the problem. That was 1965."