[std-interval] More on interval computations as proofs

[Guillaume Melquiond]

Le vendredi 22 septembre 2006 à 06:41 -0500, R. Baker Kearfott a écrit :

> In the above rewording what does "close to" mean in this context?  Also,
> "is" does not sound very normative, and I assume you want it to be normative.
> How is the following rewording?
> 
> "When \texttt{x} is a bounded interval, the result will be an approximation
>         to the real number $\frac{1}{2}(\texttt{inf(x)} +
>         \texttt{sup(x)})$."
> 
> ??

I have used "close" to mean "approximation" in my sentence. Your wording
looks less confusing to me.

> Of course, the above would be correct if midpoint were intended to be
> a of real type, but I see you have it is of <class T>.

Type T is the type of "real" numbers (intervals are of type
interval<T>). So this really is the usual definition of midpoint: a
single value, not an enclosure.

Guillaume