[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

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



Funny thing, I am in full agreement with the text of the motion, which I
repeat here:

Motion: Make the draft consistent in distinguishing between numbers
        and non-numbers (NaNs).

My amendment addresses the proposed solution:
 ... and that we adopt the term "value" to refer to
     the collection consisting of numbers and NaNs.

The proposed replacement is:
 ... and that we adopt the term "floating-point number" to
     refer to the collection consisting of numbers and NaNs.

This is in fact the predominant use in the current draft, but there is
one exception (which I propose to correct below), and an explanation is
needed for various qualifications of the phrase "floating-point number",
which I will also provide below.

There is also the issue of what to do with the many instances of the
word "number" when not qualified at all, or when the qualification is
confusing.  The qualification "floating-point number" should however
be unambiguous if this amended motion is adopted.

I hoped to remember another common term whose plain meaning is actually
excluded from a certain qualified meaning -- the primary objection to
NaNs being "floating-point numbers" -- but it escapes me now.

Text:
-----

   Append to 3.2.16, glossary entry for "floating-point number",
   the following paragraph:

      Floating-point numbers include entities called "Not-a-Number"
      (see 3.2.22), unless further qualified by a term such as
      "finite", "non-zero", "normal" or "subnormal" which imply a
      numeric entity.  A qualifier of "binary" or "decimal" does
      not exclude NaNs.


   Near the end of section 7.12.3, in the paragraph describing the
   monotonicity property,
      change "conversions" to "numeric conversions".

   (That should make it clear that the context is restricted to numeric
   entities, even though the term "internal floating-point number" is
   usually meant to include NaNs and Infinities.)


Michel.

P.S.  I will look at all unqualified instances of "number" and see if
      I can make a consistent proposal by tomorrow; if so, perhaps that
      could be appended to this amendment.

P.P.S.  In my previous note I maligned the strike-out of the word
      "normally" in 3.2.13.  I take that back, as I see that the word
      "finite" had been added (but in a colour that did not stand out).
      I now remember commenting on "an finite -> a finite" in StyleRev!
Sent: 2006-09-05 22:52:22 UTC

754 | revision | FAQ | references | list archive