Re: Motion P1788/M007.01_NaI: Discussion period begins
On 2009-08-10 07:10:30 -0700, Dan Zuras Intervals wrote:
> Again you are correct: I want more than correctness.
> I want the customer to KNOW that everything in the code
> that is the responsibility of 1788 is correct. I want
> to make sure that any bugs that are found in it are the
> responsibility of the programmer NOT 1788.
If the programmer uses an illegal construction, that's the
responsibility of the programmer.
> > If you want to teach the public that there is only one possible
> > answer, you are lying to them.
>
> If you believe that is the only possible answer then you
> believe that 1788 cannot standardize an arithmetic that
> gives provably & believably correct answers.
I don't see why. Interval arithmetic with accurate/valid mode won't
give a unique answer on all platforms, but the intervals that will
be returned will be correct (they contain the exact value). And the
accuracy may be sufficient.
Think about something like sin([x,x]) for huge x. What would the
users want?
--
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.org/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.org/blog/>
Work: CR INRIA - computer arithmetic / Arenaire project (LIP, ENS-Lyon)