[cfp-interest 3846] Re: Annex F - Special Cases & Introduction - Updates against latest C2Y

[Damian McGuckin]

On Thu, 2 Apr 2026, Jim Thomas wrote:

> Throughout, verb tenses are mixed when referring to the current draft 
> and the suggested changes. Suggest: Use present tense for current draft 
> and refer to suggested changes as ?suggested? (either explicitly or by 
> context).

Silly me.

> ?A phrase or sentence with the same mathematical meaning is often 
> written differently.? I don?t think this says what you mean. Maybe: ?The 
> same mathematical meaning is often expressed with different wording.?

Like it.

> Suggest: Avoid ?standardize?. I think that should refer to incorporating 
> syntax or semantics into the standard. After the given definition it?s 
> only used once, in the next sentence.

OK.

> What does ?tighter? mean? Suggest delete.

OK.

> ?expression, Where? -> ?expression. Where?

My typo.

> ?Where there are two tightly related special cases, one odd and the 
> other even, the domains always appear in that order.? I don?t think a 
> reader could know what this means without context. Suggest delete.

OK.

> ?less impactful consistencies? -> ?less impactful inconsistencies??

Got it.

> Phrases like ?what should be?, ?often to the point of confusion?, ?much 
> clearer?, ?the confusing issues?, and ?impactful? seem like judgement or 
> speculation, since no evidence is given of users being confused or 
> impacted.

OK.

> ?The elimination of current rounding direction has required several 
> kinds of changes. For now, those changes are in Annex F.? Suggest: ?This 
> proposal eliminates use of the term current rounding direction from 
> Annex F.?

OK.

> The part in Rationale about current rounding direction is the only 
> actual background item. Suggest combining Background and Rationale into 
> Background/Rationale or Introduction.

Sure.

> ?The special case? -> ?A special case?.

Yes.

> Suggest: ?A signaling NaN input to scalbn() in F.10.4.19 mistakenly 
> infers that a signaling NaN input needs ?? -> ?F.10.4.19 mistakenly 
> implies that a signaling NaN input to scalbn needs ??.

OK.

> ?Behaviour? -> ?behavior?

Sorry. I will correct it to American spelling. In most places I got it 
right.

I notice that the standard uses correct English with the noun "practice". 
Good stuff.

> Suggest: ?i.e. that a quiet NaN be returned for signaling NaN input, the 
> scenario now inferred by the revised F.10.4.19 which explicitly excludes 
> signaling NaN input? -> ?i.e. that a quiet NaN be returned for signaling 
> NaN input, which can now be inferred by the revised F.10.4.19?.
>
> ?for all x (including NaN)? does not include signaling NaN input (given 
> our convention for the meaning of unqualified ?NaN? in Annex F), but 
> does not explicitly exclude it. It?s ironic that we?re excluding 
> signaling NaNs by adding ?including NaN? ...

A long while ago I was going to propose

 	-OO <= x <= +OO U x is a quiet NaN

where OO was an Infinity symbol and U was a logical union sign.  As you 
say, there are only a few places where it is needed.

> I hate to bring this up, but maybe for these special cases we should 
> consider changing all ?(including NaN)? to ?(including quiet NaN)?. 
> There are only six cases.

No problems. I was thinking the same thing.

In the changes to F.2.2 Infinities and NaN, I did exactly that and
spelt out quiet NaN all the time.

Thanks - Damian