Re: Motion P1788/M0021.2:IntervalOverlapping NO
On 2010-12-12 16:52:07 -0700, Alan Eliasen wrote:
> I vote NO on M0021.2.
>
> I do not believe that the set of operations listed in this paper are
> generally useful, and it enlarges the specification without clear
> benefit. There is nothing in the paper that could not be obtained by
> allowing access to the infimum and supremum of each interval and
> allowing the user to make their own comparisons.
This is also true for Motion 13.04, which has passed (I note that
you also voted NO on it, though, but this wasn't such a firm NO).
BTW, a point that Motion 13.04 is missing is set-based definitions,
I mean the equivalent of Table 1 of Motion 21.2.
> In addition, I think some of the comparisons are actively dangerous,
> such as "meets" or "starts" or "finishes" or "finishedBy" or "startedBy"
> or "metBy", or "equalP". These essentially require testing equality of
> two imprecise floating-point numbers, and doing something based on that
> test.
This is also true for Motion 13.04 (which has a = b, but also
two different "a precedes or touches b" and "a precedes b").
Anyway if you want to avoid such problems, you should define
an arithmetic without discontinuous functions, thus without
comparisons.
> This is almost always bad practice, and usually indicates
> misunderstanding of uncertainties in floating-point math.
Note that IEEE 754 offers some guarantee.
--
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / Arénaire project (LIP, ENS-Lyon)