[std-interval] Revised document available

[Dr John Pryce]

Guillaume and all

At 13:20 21/09/06, Guillaume Melquiond wrote:
>This may be a bit verbose...Could we not get the same effect by simply saying
>something along the lines of the following in place of 26.6#3?
>
>        Intervals are connected subsets of the set of real numbers.
>        Which subsets are representable by \texttt{interval<T>} is
>        implementation- defined. An implementation shall support at
>        least the empty set $\emptyset$, the whole set of real numbers
>        $\mathbb{R}$, and any singleton interval $\{x\}$ for $x$ a real
>        number representable by a floating-point number of type
>        \texttt{T}.
>
>So it leaves to the implementation any consideration about closed/open
>intervals. As only singleton intervals are mandated, a midpoint-radius
>representation will not be prevented. It might even accommodate a cset
>interpretation if we are careful enough with the first sentence. I don't
>think operations have to be defined here, as they are (almost all)
>defined by expressions on sets.

Yes, my offering was a bit verbose. I would be happy to go along with 
your words for now. They seem to say what is essential and have the 
singleton clause to tie the precision of interval<T> loosely to that 
of T. Any objections I might think of would be very minor.

Regards

John
Dr John and Mrs Kate Pryce
142 Kingshill Rd
Swindon, Wiltshire SN1 4LW
UK
Tel (+44)1793-331062