From The Register:
"Invariably someone puts a flight plan wrong and it borks* the system," one source told El Reg on condition of anonymity.
That should never happen. And (hopefully) the bits that fly the planes are built to a higher standard. One would expect a well written application to skip over the bad data, raise an alarm, log it, but keep going. Something like dividing by zero (
Sunk by Windows NT one of my favorite SNAFU examples of poor system implementation) should never lock up an operating system. In a well written app, it shouldn't even slow down processing of the remaining good data.
Part of the problem is cultural. The people who do mainframes have historically guarded their domain from systems engineers and real time software experts. For people who did things like payroll systems, it was acceptable to print the crash report, go through the data and re-punch the defective card. And then run it again. That's just not going to cut it in real time.
*And quit picking on poor Judge Bork.