An Oops! in Level 2 text
Vladik, P1788
Really someone should have pulled me up short about wid(xx) (width of an interval) in the Level 2 draft text. I had written
wid(xx) returns the same value as 2 ∗ rad(xx)
This is quite wrong, e.g. in 3-digit decimal, if xx = [.234,.235] then mid(xx), if we follow the roundTiesToEven rule of motion 37, will be m=.234, and rad(xx) will be .00100. My recipe above gives width w=.00200 but surely the only sensible result is .00100, the actual width.
So IMO the correct recipe is
wid(xx) returns the smallest F-number ≥ the mathematical width xhi-xlo.
Vladik: what is the Level 2 width of Empty? I would be happy with
NaN, on the grounds that it is genuinely undefined;
or
-oo, on the grounds that wid(xx) = (sup xx) - (inf xx), and when
xx=Empty, one has sup xx = -oo and inf xx = +oo.
John Pryce