From: Remi on
Hi to all,

I'm a mechanical engineeing student and for the past 2 months, I build a Mathlab GUI who does a bunch of data analysis, plotting, etc at my job. My boss asked me to write a software verification report since the software's use is critical (spacecraft thingy...) but I have no clue how to do this.

What should be included in a software verification/certification(?) report?

TIA!
From: Walter Roberson on
Remi wrote:

> I'm a mechanical engineeing student and for the past 2 months, I build a
> Mathlab GUI who does a bunch of data analysis, plotting, etc at my job.
> My boss asked me to write a software verification report since the
> software's use is critical (spacecraft thingy...) but I have no clue
> how to do this.


> What should be included in a software verification/certification(?) report?

No offense intended, but considering the context, that sort of thing really
needs to be left to a team of professionals -- a _team_ to go over the code
line by line and think of all the different ways the code might go wrong;
someone to write a test framework; one (or better, several) -different- people
to write test-cases (someone who doesn't know the details of how the test
framework works, since people who program algorithms have difficulty in seeing
how the algorithm might fail); someone who specializes in numeric analysis to
go over the algorithms with a fine-tooth comb and figure out what the limits
of accuracy are -- and, of course, someone who knows Matlab well to catch the
ways you might be misusing a call or operation...

Remember, just a few years ago the spacecraft that was lost because one team
"knew" that everything was programmed in metric, and the other team" knew that
it was programmed in Imperial.

For proper Software Verification Reports for something critical, you need the
kind of Bloody Idiot who says, "Hmmm, I wonder what happens if I enter
system("format C:/")
as the response for the km/s prompt, or over in this numeric edit box?"
From: ImageAnalyst on
Remi :
You may be interested in the guidance the US Food and Drug
Administration gives in how to validate software:
http://www.fda.gov/downloads/MedicalDevices/DeviceRegulationandGuidance/GuidanceDocuments/UCM085371.pdf
Good luck with that,
ImageAnalyst
From: Remi on
I totally agree with you about this. I think I didn't made myself clear (english isn't my native language). The GUI I'm working on will be use internally by my boss and me only, so we do know how it works. I'm not trying to write a "bulletproof" software against idiots.

The verification report is more about the math behind the software than the software itself, to insure that the displayed numbers are the good one.
From: Remi on
ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <b242bfde-8beb-420f-acfd-a50eda52be04(a)b30g2000yqd.googlegroups.com>...
> Remi :
> You may be interested in the guidance the US Food and Drug
> Administration gives in how to validate software:
> http://www.fda.gov/downloads/MedicalDevices/DeviceRegulationandGuidance/GuidanceDocuments/UCM085371.pdf
> Good luck with that,
> ImageAnalyst

Thanks, I'll take a look!