Re: Sweden Fighter Crash

[skates.gsfc.nasa.gov!bambam.gsfc.nasa.gov!nbssal@ames.arc.nasa.gov (Step he Leake)]

In article <26347@alice.att.com>, peju@alice.att.com (Peter Juhl) writes...
> 
>Last week Bjarne Stroustrup posted something about the fighter
>that crashed in Sweden. Eventhough his main point was about
>the difficulty of comparing languages, and the traps that people
>fall into when they do so, there was some follow-ups about the
>crash as such.
> 
>Somebody from Sweden expressed doubts about Bjarne knowing anything about
>what caused the failure, which Bjarne attributed to software.
> 
>This just came in on the apnews
> 
>	   STOCKHOLM, Sweden (AP) _ Sweden's new jet fighter crashed during
>an air show last week because of a computer flaw, an inquiry board
>said Wednesday in clearing the pilot of blame.
>	   Producers and military officials had been aware that vigorous
>movement of the control stick could cause the JAS 39 Gripen to go
>out of control because of computer overcompensation. But they
>underestimated the problem, the government board said in a
>preliminary report.
> 
>[ ... stuff deleted ... ]
> 

This sounds more like a bad value for a gain parameter, causing the control
system to go unstable, rather than a flaw in the program or compiler. I work
with real-time control systems; there are  many things that can go wrong even
after the compiler and algorithm bugs are all found!

If it is a bad gain, then it is the control engineers, not the software enginee
rs,
who are at fault. Let's not assume everything is a software fault just because
we are software people!

Stephen Leake	NASA Goddard Robotics Lab
internet : nbssal@robots.gsfc.nasa.gov