Hacker News Reader

Comment by robryan | Hacker News Reader

It's interesting just how many different systems needed to be considered by the pilots after the incident. I've read articles before about how a pilots routine job can be very boring but it's good to know they have built up all that experience and training for moments like these.

View on HN

What strikes me is that the software he's talking about appears to be badly designed; or at least, not well designed for this rare situation where you've got dozens of error messages coming up at once and they're not prioritised.

Also,

...once we’d extended the undercarriage using the alternate system we had no indication it was down until we’d gone to the system page to make confirmation of that.

(What, there's no big-ass mechanical dial with "Gear up" and "Gear down" on it?)

and

In the Airbus and the A380 we don’t carry performance and landing charts, we have a performance application. Putting in the ten items affecting landing performance on the initial pass, the computation failed. It gave a message saying it was unable to calculate that many failures

I think this is one of the primary religious debates between Airbus and Boeing pilots - Airbus is a much more highly computerized experience inside the cockpit vs. Boeing which is still a more mechanical/tactile experience. For some fascinating background, see the NOAA on the recent AirFrance crash. They cover this issue specifically and essentially concluded that the pilots were likely overwhelmed with the number of error messages and warnings coming at them from the plane's computers and it became paralysis by analysis which they ultimately couldn't recover from.

Paralysis by analysis is one of the most common ways more intelligent opponents are beaten by less intelligent ones. From what I understand, this seems to hold true in chess, warfare, and a lot of other contexts. This might be good for us to remember in the coming years.

Yeah, bad software is the point Philip Greenspun makes about the incident: http://blogs.law.harvard.edu/philg/2010/12/11/airbus-a380-en...

It is usually a trait of critical systems, imagine a similar disaster on the space shuttle, while being exceptionally reliable software I would imagine the same kind of issues with ease of use. Although you may have more mechanical control there.

This is actually being recognized now and being addressed. The thing is flight software has to conform to the DO-178B standard which is antiquated in how you can verify systems. Everything has to be done manually, even model/computer generated code. The result is you can handle far fewer conditions/branches than you'd expect.

That is changing with DO-178C which allows model generated code as long as you verify the modeler system. Then you'll be able to put in all the conditions/priorities and have a model spit out to be included in the avionics software.

As to the boeing vs airbus argument? As a private pilot I like a hybrid. I want the physical linkage, but I'd prefer if my systems management be fully computerized. There's just no reason for my gps/flight computer to not tell me "Bingo destination!" when I've got just enough fuel to get to my destination (+1hour emergency/maneuvering required by law).

DO-178C opens up a HUGE green field for avionics development. It'll be cool to see what happens in the future.

Reply on news.ycombinator.com