Gopher words

[Mike Cowlishaw <MFC@xxxxxxxxxx>]

Looking at my to-dos from last night, I see all of them involve
'gopher words' -- words which we try and avoid using in the draft
because even if we give them a formal Definition for 754r they
are still likely to be confusing or mis-read.  They tend to keep
popping up, and often cause trouble...

Here's a list of the main ones that come to mind, together with
comments and the result of global search for each in the Aug. 18
draft.

* assignment -- we use copy (to a destination) for this

    This only appears in section 7.8 para 5 (once, p36) and in
    B.3 Assignments (p51).

    Suggestions:

      In 7.8: change 'for assignments' to '(including conversions
              during a copy)'

      In B.3: either use the phrase 'copy to a destination'
              throughout the section (and heading), or define
              what we mean by assignment in the first sentence of
              the section (this one's informative anyway).

* number (as in floating-point number, etc.)

     This is used 152 times and is used in many different forms.
     I started going through but typing detailed notes got too
     tedious.  Here's a summary:

     The phrases 'finite number', 'normal number', 'subnormal
     number', 'real number', 'the number of' (digits, etc.) are
     common.  These forms are all clear enough and well-defined.

     There are, however a variety of other forms:

       'representable number' (in a few places)
       'number' (in 1.3 Exclusions, p9, and many other places)
       'floating-point number' (in Scope and many other places)

     In many cases, these refer to the full set of representable
     entities (including NaN and Infinity), and in (roughly) as
     many other cases these are just referring to finite numbers.

     I suggest that all the cases where they just refer to finite
     numbers that the word 'finite' be added (often replacing
     'floating-point', I think).

     That leaves the cases where these terms refer to the full
     set.  We could:

       a) Use 'floating-point number' for this

       b) Make up a new term for the concept and stop using
          'floating-point number' entirely

       c) Add a clarifying word (e.g., 'floating-point number
          set')

    Needs a Style Review discussion, I think, before starting to
    make any specific changes, as there are so many places to
    inspect.

* nonnumber

    This is used only once (in 1.2 Exclusions, final bullet
    (p9)).  I suggest we rephrase that one (it is not clear what
    it is meant to mean, anyway).

* precision

    This has at least three common meanings in computing, see:

      http://www2.hursley.ibm.com/decimal/decifaq2.html#precision

    It appears 32 times in the document, but we are in much
    better shape with this one: in all cases (from a quick skim)
    it has the meaning as defined for formats in section 5.3 (p17):

      p = the number of significant digits (precision)

    with the implication that this is also the rounding precision
    (there are some uses of 'infinite precision' but these are
    clear enough).

    Suggestion:  add a new Definition to section 3, perhaps:

      3.2.x precision:  the number of significant digits that can
      be represented in a format, or the number of digits to
      which a result is rounded

* program

    Many uses are in informative text such as the Introduction in
    a non-specific way, which is OK.  Similar 'outside the
    standard' use in 3.2.38 User.

    However it is used quite widely elsewhere (and in the phrase
    'program text') but not, I think, in a way that it really
    changes the intended meaning if you have different
    interpretations of the word.

    So, I think I agree with Jim on this one, now -- probably OK
    to leave as-is.

    (Note to Editor: 'programming' is misspelled 'programing' in
    at least one place, worth doing a search.)

* value

    114 uses.  This one has been a problem in the past, related
    to 'number' (can one have a 'value' of NaN)?

    I haven't looked at all 114 ... but it looks as though the
    only problematic uses are when we talk about 'the value of' a
    'number'.   I think these will clean up automatically when we
    clean up 'number'.  Maybe worth checking/reviewing after
    that.

Mike