Motion P1788/M007.01_NaI NO

[Guillaume Melquiond <guillaume.melquiond@xxxxxxxx>]

I vote NO.

I have some mixed feelings about this motion. While I can see the 
usefulness of NaI, I don't think introducing it just to represent the 
result of [3,NaN] is a reason compelling enough. Having an exception 
(IEEE-754 meaning) with empty as a default behavior would work just as well.

I also don't approve of the unicity of NaI. As soon as we introduce it, 
we may just as well follow 754's lead and allow payloads that propagate 
along computations in order to differentiate between the various issues 
(uninitialized data, invalid data, failed construction, etc).

Best regards,

Guillaume