Re: Boeing and Dreamliner

[Hyman Rosen <hyrosen@mail.com>]

Mike Silva wrote:
> Good heavens, no, you're thinking of C!  There was no buffer overflow,
> and there was no bug.

The people who coded the Ariane 4 SRI never described, in
their program or in their documentation, the fact that if
the Horizontal Bias became too large, their alignment code
would blow up. The Ariane 5 people would have had no reason
to believe that they were getting code that depended on a
restricted trajectory to be valid. Why should they? The SRI
didn't even do anything useful after takeoff. It was a
perfectly natural assumption that they were getting SRI
software, not SRI-for-Ariane 4 software. Have a look at the
report - <http://java.sun.com/people/jag/Ariane5.html>.

Of course, these kinds of errors can happen in any language.
And that's the point. When certain people start claiming that
using C++ is an actionable offense, the Ariane 5 case shows
that using Ada is no panacea.

And just to make the buffer overflow analogy again, since you
don't seem to get it, what is a buffer overflow? It happens
when a programmer makes an assumption about the size of the
input he expects his program to receive, and circumstances
cause that size to be exceeded, whereupon the program functions
erroneously. Did that not happen in the Ariane 5 case? Is there
a statue of limitations on buffer overflows, so that if it takes
a long time for them to show up, the original programmer is not
to blame? If the programmer insists that the unchecked size
limits are needed, for efficiency, say, shouldn't the limits be
documented?