Airbus is not going to gloat too much because
All this shows is that those who write safety critical software are not supermen, and are probably no better than your nearest PHP hacker who has left SQL injection back doors all over the place
Probably the biggest thing that protects us is the fact that most of the systems are really old, perform very narrow functions (basically do the same thing every day; airline pilots tend to press the same buttons in the same order every time), and a lot of them (like jet engine ECUs) are carefully designed self contained boxes and then don’t get changed for many years.
Defects discovered by crew are reported internally, covered up immediately because the Daily Trash sharks are always circling, while the mfg sh**s itself and issues a rapid fix under the guise of an innocent sounding “feature update” so the certification agency never finds out
One might think the reason Airbus are not making political capital out of the B737 saga is because they are upright and decent people, but in fact they are keeping quiet because they know their cupboard has just as many skeletons in it as all the others.
Many years ago, 1983, I designed a multizone heating controller which had self-learning optimisation on the advance turn-on. It was all written in Z80 assembler. Many many hundreds were sold and installed (c. £500 each) and no bugs were ever found. Well, all of them will have stopped working in 2012. The reason is that it computed the day of the week from the date (which came out of a clock chip) and the algorithm I used was just a lookup table which took advantage of the fact (?) that a calendar repeats every 29 years, and I never provided for any subsequent 29 year period. I mean, it’s obvious, in 1983 I was 26, and when you are 26 it is unthinkable that you will ever be 55 because nobody who is of any interest to a 26 year old is that old. I left the company (my first business) in 1991, and it went bust in 1993. Go figure, as they say…
I recently spoke to someone who worked on the Honeywell KFC225 autopilot, introduced into general aviation c. 2000 and dropped c. 2003 due to widespread servo burnout issues caused by defective autopilot computer software (believed to be integer maths underflow / dramatic loss of precision). The software had one file per function, because the FAA would push for recertification if more than x functions were changed. If you had 100 functions in one file then any edit of that file changed all 100 functions. The result was an almost unreadable source code…
Home
Help
Search
Login
Register
