RE: Amendment to motion addressing the need for consistency in distinguishing between numbers and NaNs

[Michel Hack (1-914-784-7648) <hack@xxxxxxxxxxxxxx>]

I understand your discomfort, and partially share it.  I was however
more uncomfortable with the choice "floating-point value", because
the word "value" appears so many times in the document in a context
where it implies "numerical value", even though I agree that that
meaning is not inherent in the word "value".  Especially problematic
is the dual description of encodings in terms of the value v and the
representation r.

We have since heard a third proposition (though not formally), namely
"floating-point datum".  It at least has the merit of not conflicting
with any prior use: "datum" is not used yet, and "data" appears in
contexts that do not conflict with an explicit new technical definition.
What remains to be seen is how the text will read when all instances of
"floating-point number" (all instances of "Nancy", if you wish) are
replaced accordingly (including those I introduced in my clarification
for previously unqualified instances of "number").

Btw, what I meant by "wishful thinking" was the fact that the gruntwork
of identifying all instances that need to be fixed, and especially the
consequences on other instances of "value", still needed to be done.
I'm not entirely sure I agree with the notion "pick a name first, then
worry about fixing the document", because some choices may require much
more severe surgery than others.  I felt that "floating-point number"
required the least amount of surgery; in fact, I already provided the
necessary suggestions.

I wonder whether you could try to suggest more explicit ways to deal
with all (50 or so) instances of "value" in the current draft, so that
we have an estimate on the amount of surgery required.

I also wonder whether Hossam Fahmy could try to see how the replacement
of "floating-point number/numbers" with "floating-point datum/data"
would work out, taking my suggested clarifications for plain "number"
into account.

Michel.

Sent: 2006-09-06 16:39:59 UTC