Subject: Re: rocket science
From: "Forrest J. Cavalier III" <mibsoft@mibsoftware.com>
Date: Mon, 28 Feb 2005 13:05:01 -0500

Tom Lord wrote:

> I believe that your analysis contains beginner's mistakes in how you
> reason about the structure, function, and risks of computing systems.
> The RISKS mailing list is a good source for learning about such
> things.

That doesn't show much evidence that you practiced much tact in your
time off-list, Tom.  <stephen@xemacs.org> can defend himself well
enough, but gosh you are, um, annoying to call a clueful frequent poster
to fsb a beginner.

I'd say that since most of the gloom and doom posted to the RISKS mailing
list never causes the large scale economy crippling predicted to be
possible, that means it is not a good source for learning about what
actually happens, or is even likely to happen.

Economics and business are about what actually happen, not what should
happen, or what might happen.

Anything NASA does related to shuttle software is a pretty bad example
for anything I'd ever be interested in.  I don't even know that the
shuttle software has even been that well tested.  The NASA shuttle bunch
lives on making everything happen "just so" and getting just
a little outside that things go wrong quickly, with two spectacular
disasters in a relatively small number of flights.  The Challenger
investigation board arrived at a 1/130 chance of full shuttle and
crew loss per launch.  Looks to me that that average is a pretty
good predictor.

The MARS rover software successes seem to be more based on sending up a
flexible platform and having clever people on the ground adapt and
debug as things come up.  That's a smart strategy, but seems to be the same
old, same old: release known buggy code early and fix it later.

Programmers have been doing that from the beginning, so reprogrammable
flash memory has been saving a lot of marketing butts lately and
will continue to do so.  That's what the marketing guys expect.

Marketing runs the world's commerce, not programming.