Re: An attack on a mint

Bill Frantz wrote:
>
> Take for example the code for VatTP in E. The unit tests for that
> code that I used while developing it were never added to any source
> repository. Along with the machine they ran on, they are gone. It's
> not a matter of just not being run. They are gone.
>
My Python and Java codes have the unit tests in the same source file as the code that does the work.  That makes it hard to lose them and easy to update when you want to add a test.  Since I learned to avoid loops in my object graphs, I only need to go through the tests in the graph once.
>
> I think before you consider refactoring high reliance code, you
> need to consider the cost of re-certifying it. If you were working
> on Shuttle code for NASA, those costs might be prohibitive.
>
Some folks a GE Nuclear told me that they could accept any change that produced bit for bit the same 200 MB of output that had been certified.  If one bit changed, it would cost them hundreds of thousands of dollars, and those were 1980s dollars.

________________________
Alan Karp
Principal Scientist
Virus Safe Computing Initiative
Hewlett-Packard Laboratories
1501 Page Mill Road
Palo Alto, CA 94304
(650) 857-3967, fax (650) 857-7029
http://www.hpl.hp.com/personal/Alan_Karp