Re: Empty interval representations & Motion 13...

[Dan Zuras Intervals <intervals08@xxxxxxxxxxxxxx>]

> Date: Thu, 22 Apr 2010 18:17:13 +0200
> From: =?ISO-8859-15?Q?J=FCrgen_Wolff_von_Gudenberg?=
>  <wolff@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
> To: Dan Zuras Intervals <intervals08@xxxxxxxxxxxxxx>
> CC: stds-1788@xxxxxxxxxxxxxxxxx
> Subject: Re: Empty interval representations & Motion 13...
> 
> Dan
>    your representation may simplify comparisons but on the other hand
>    it complicates arithmetic.
> Juergen
> 

	How so?

	I realize that it may be necessary to test for empty
	before it participates in a numeric computation but
	how is that made more complicated when the test for
	empty is a test for [+oo,-oo] rather than for [NaN,NaN]?

	Or are you counting on the fact that most arithmetic
	functions are NaN preserving & not testing at all?

	I can see how that might work but it seems imprudent
	at best.

	Is that how empty is done today?

	If so, how is the appearance of a NaN in an interval
	distinguished from the failure of some floating-point
	computation in need of diagnosis?

	Or are they not distinguished?

	And, if they are not, how are we to assure the world
	of the correctness of our results?


				Dan