On Sep 19, 5:31 am, Bob Noel
wrote:
In article .com,
Le Chaud Lapin wrote:
Hi All,
[snip]]
and cost difference
should be huge . [Yes, I know, reliability, FAA
certification...yada...]

You say you know about reliability, etc. But do you really know what it takes
to do the safety analysis?

Nope. I just know that it will be a fixed cost. My guess is that it
would be under $100,000,000. If so, then those costs would be
recuperated.

What are the failure modes of these components?

Same as for most pieces of software and hardware.

How will failures
and errors be detected and handled?

Self-checking, pre-flight, and during flight, redundancy, etc.

How will component changes
be handled?

With more professionalism than the free pop-up blockers, for example.
The first time a plane crashes due to a company's gross oversight
(read, bad engineering), they would get license revoked by FAA. Also,
the components would still have to be checked.

How much will it cost to repeat the appropriate analyses
when various vendors roll part numbers?

Dunno...I think this is the crux of the issue. The existing older
components are well understood and familiar. 5,000 lines of C code is
not as familiar.

How will you determine that
the part hasn't changed when the vendor didn't change the part number?

Abstraction barrier. The component would have to comform to
specification. After that, they can changes as they wish.

(Don't laugh, I've seen an LRU no longer work in a particular aircraft when
a chipset vendor changed a production process which ever so slightly
changed functionality but the vendor didn't change the part number).

And do you have any concept of what it would take to put a commodity
OS like windows into a safety-critical application?

From a technical point of view, I guess, yes. From a "How much must I
pay the FAA and fight political fall-out" point of view, no.

-Le Chaud Lapin-