Re: You need not say too much, was: Re: Verifying Chipmunk intervals
> Subject: Re: You need not say too much, was: Re: Verifying Chipmunk intervals
> From: John Pryce <j.d.pryce@xxxxxxxxxxxx>
> Date: Tue, 6 Sep 2011 23:03:53 +0100
> To: stds-1788 <stds-1788@xxxxxxxxxxxxxxxxx>
>
> Dan
>
> On 6 Sep 2011, at 18:39, Dan Zuras Intervals wrote:
> >> And could I run tests to check the following, which is IMO an important
> >> property of a representation?
> >> ss=w(xx) implies xx=r(ss),
> >> whence distinct internal intervals xx have distinct representations ss.
> >>
> >> Seeking your opinions.
> >
> > This may be an example where, by specifying too much,
> > we get ourselves into trouble with other standards.
> >
> > How about saying "It shall be possible" to interconvert
> > such that the above is true & leave it at that?
>
>
> But the provider of an exotic interval type might be deceitful,
> or have failed to notice a bug. I think Michel has diagnosed my
> worries correctly.
>
> John
Agreed.
And tests must be written.
But neither over specification nor testing will
assure us that we have guarenteed the correctness
of any given implementation.
Especially if deceit is considered a possibility.
I'm not sure proofs are available at this level.
There are those among you that can address this
issue better than I can.
Can one specify a provable implementation?
Dan