Ed Kyle ) wrote:

: Thomas Lee Elifritz wrote:
: Ed Kyle wrote:
:
: The reported facts about DART are consistent with a software
: issue of some kind. Software fails when it has not been
: adequately tested.
:
: No, software usually continues to execute, or executes to completion,
: in which case the end state may or may not result in a lock up,
: or an infinite loop.
:
: Usually it's the software engineers that fail.

: Program managers who fail to test to find faults in
: complex software are the ones who should be blamed.

Blame alone won't fix the problem nor will it prevent similar problems in
the future.

: No software engineer can design perfect algorithims that
: work perfectly in every circumstance, especially when
: the circumstance is rangefinding and/or pattern recognition
: in low earth orbit using detectors and guidance systems
: that may have originally been designed for other purposes
: and were brought together in a hurry on a limited budget
: and tight schedule.

Right, interface problems between designed modules do occur with badly
designed software, but someone must see the big picture and they should be
involed with the testing, at least on defining the expected results.

Personally, this is why I like the Rapid Prototyping Model rather than the
Waterfall Model when developing SW. The former allows for problems to be
worked out early and if done right mitigate risk early on as well. The
latter can have it be months or even years before the first 'drop' where
anyone can see results and by then a lot of work could be in something
that is basically bad. Then comes the work of trying to fix rather than
redo. Rapid Protyping or the Spiral Model allows for less redo which tends
to save time and make a better product.

Eric

: - Ed Kyle