Letter in local paper - Speedophiles

[Ian Smith]

On Sun, 7 Sep 2003 23:10:52 +0100, Tony Raven <[email protected]> wrote:
>
> "Ian Smith" <[email protected]> wrote in message
> news:[email protected]...
> > On 8 Sep 2003 05:28:29 +0950, dailuggs <[email protected]>
> wrote:
> >
> >
> > Except that requires putting your faith in programmers. At least the pilot (most of the time)
> > has a personal interest in not hitting the ground - the programmer is probably thousands of
> > miles away, tucked up in bed. It's not just idle speculation - as anyone that knows what a
> > boeing chinook fadec is can tell you (cue Paul McCartney song).
>
> Yes but they only have to get it right once, pilots have to get it right every time.

But they have to consider every eventuality, and every combination of every eventuality, in advance.
Pilots only have to consider what's best under a single existing set of conditions.

regards, Ian SMith
--
|\ /| no .sig
|o o|
|/ \|

 

[Peter B]

"Ian Smith" <[email protected]> wrote in message news:[email protected]...
> On the bright side, Microsoft has mostly stayed out of embedded systems. Sadly, it probably
> won't last.

Not quite the same I know but didn't a US naval vessel some years ago become useless in the Gulf
after the uncrashable NT it relied upon crashed? (alegedly).

Pete

 

[Pete Barrett]

On Mon, 8 Sep 2003 14:27:55 +0100, "Tony Raven" <[email protected]> wrote:

>Still ignoring the fact that 80% of air accidents are due to pilot error I see ;-)

What proportion of flight time is under automatic systems, and what under manual? More
significantly, what proportion of the most dangerous times in a flight (take off, landing, bad
weather) is on autopilot, and what on manual?

Pete Barrett

 

[David Nutter]

Ian Smith <[email protected]> said:

> But they have to consider every eventuality, and every combination of every eventuality, in
> advance.

Sadly, exhaustive testing is wholly impractical for even fairly trivial programs since it takes far
too long (several ages of the Universe for example). Safety critical systems development often makes
use of formal methods to prove the correctness of a program or part of a program. However, this
elevates the possibility of error to the specification level and is quite difficult in its own
right. AFAIK Robert Glass's "Challenge to the Formal Methods Community"[1] has never been answered.

Even assuming perfect program design and implementation you can never know that you've forseen every
situation your system will have to deal with.

> Pilots only have to consider what's best under a single existing set of conditions.

Funnily enough, that's how most software engineers work: "We have (very specific) problem X, let us
now create a solution for it". Along the way, you generally find out that problem X is not what you
thought it was and your solution ends up far more complicated than you anticipated. Oh, and your
customers want to apply your solution to a totally different problem
Y...

The RISKS digest is a good starting place for anyone wishing to be convinced that computer systems
are a menace!

http://catless.ncl.ac.uk/Risks

That said, when you get it right software is brilliant: much better than filthy fallible humans for
repetitive tasks.

ObBicycle: Bleedin' Shimano. Not only have they sent my Nexus back to my LBS in pieces, they have
neglected to include one of the bearing cones. Grr...

Regards,

-david, embittered programmer

[1] Namely, show unequivocally that they've achieved something useful with proofs of
correctness etc.

 

[Ian Smith]

On Mon, 8 Sep 2003 18:28:10 +0000 (UTC), Peter B <[email protected]> wrote:
>
> "Ian Smith" <[email protected]> wrote in message
> news:[email protected]...
> > On the bright side, Microsoft has mostly stayed out of embedded systems. Sadly, it probably
> > won't last.
>
> Not quite the same I know but didn't a US naval vessel some years ago become useless in the Gulf
> after the uncrashable NT it relied upon crashed? (alegedly).

The Yorktown, I think. I don't recall where it was when it was supposed to have happened.

regards, Ian SMith
--
|\ /| no .sig
|o o|
|/ \|

 

[Tony Raven]

David Nutter wrote:
>
> The RISKS digest is a good starting place for anyone wishing to be convinced that computer systems
> are a menace!
>
> http://catless.ncl.ac.uk/Risks
>

Bit like cycling, there are dangers but risk is very small and the overall benefit in positive in
favour of safety. Which is why they fly planes automatically in the main these days, want to
introduce automatic systems to stop train drivers making mistakes etc etc.

Tony

--
"If you tell the truth you don't have to remember anything." Mark Twain

 

[Tony Raven]

Pete Barrett wrote:
> On Mon, 8 Sep 2003 14:27:55 +0100, "Tony Raven" <[email protected]> wrote:
>
>> Still ignoring the fact that 80% of air accidents are due to pilot error I see ;-)
>
> What proportion of flight time is under automatic systems, and what under manual? More
> significantly, what proportion of the most dangerous times in a flight (take off, landing, bad
> weather) is on autopilot, and what on manual?
>
>

These days > 99% automatic, the more so in bad conditions, which makes the pilots even more
dangerous!

Tony

--
"If you tell the truth you don't have to remember anything." Mark Twain

 

[Richard Bates]

On Fri, 5 Sep 2003 09:54:12 +0100, Peter Fox <[email protected]> in
<[email protected]> wrote:

>Great new word.

Is there a difference between a "Speedophile" and a "Fastard" ?
--
If ingnorance is bliss then I am the erm er luckiest thingy in the whatchamacallit. To mail me,
change the obvious bit to richard

 

[Peter B]

"Tony Raven" <[email protected]> wrote in message
news:[email protected]...
> Still ignoring the fact that 80% of air accidents are due to pilot error I see ;-)

The M1 Kegworth crash was exacerbated, if not caused, by the pilot disbelieving his instruments and
instead reacting to what he sensed and then shutting down the wrong engine.

Mind you, he blamed it on the training he'd received, at least my fellow motorists don't do that ;-)

Pete