Re: [Fwd: motion 7 NaI]
On 2009-08-24 09:36:59 +0200, Jürgen Wolff v Gudenberg wrote:
> revised rationale is attached.
2 editorial corrections:
* datums -> data
* occurence -> occurrence
Also, even though this is just an example, I wonder whether
differencing NaI from NaN with sNaN vs qNaN is a good idea.
Is sNaN support always correct in languages?
> I know that the issue is contraversal. To come to a sound decision I
> would like to ask you to vote for one of the following three
> alternatives.
>
> 1. YES ,there should be a unique NaI
> 2. NO , there should be no NaI at all
> 3. NO , there should be a NaI with payload
It isn't very clear. For instance, in IEEE 754, the payload doesn't
have an influence on the result (except for the payload of the result
when it is a NaN); this makes the payload useless to distinguish some
different NaN types such as invalid construct or missing data. So,
what would be the payload status in P1788?
--
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.org/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.org/blog/>
Work: CR INRIA - computer arithmetic / Arénaire project (LIP, ENS-Lyon)