08-282
To: J3
From: Malcolm Cohen
Subject: 185 editorial deferral reasons
Date: 2008 October 22

1. Introduction

Meeting 185 passed a number of papers with editorial improvements and
fixes.  Some of these have been rejected for inclusion in 08-019r1
(i.e. deferred to the next meeting) as the editor considers that they have
errors or need further work.

This paper outlines the reasons for their deferral, it is not for action.
New papers containing revised edits for these issues will doubtless be
submitted for consideration at meeting 186.


2. Deferrals

  08-198r1.

    This is just wrong.  Its last sentence is nonsense on several levels.
    The existing text, duplicative though it is, is a substantial
    improvement on this.  It's bad editorially too.

    I recommend starting from scratch or abandoning the idea.

  08-203r1.

    It is unacceptable to do wholesale replacement of entire long
    paragraphs when simple insertion/deletion instructions suffice; apart
    from unnecessary work for the editor it increases the risk of typos and
    of lost or incorrect markup.  Secondly, the final instruction says to
    put it into a non-existent subclause (there is no subclause with the
    stated title).  Finally, putting it into the numbered subclause (i.e.
    ignoring the wrong title) results in an interp request since that
    subclause says it is about "block DO".

    The editor had a simpler alternative paper that somehow seems to have
    been lost, perhaps due to a glitch in the J3 server when he tried to
    submit it.

  08-209r4.

    It is unacceptable to do wholesale replacement of entire long
    paragraphs when simple insertion/deletion instructions suffice; apart
    from unnecessary work for the editor it increases the risk of typos and
    of lost or incorrect markup.  Secondly, the replacement text is not
    grammatical; the existing text is an improvement on it.  (I agree with
    the comment that the existing text needs improvement, but this is not.)

  08-216r2.

     Nice try, but contradicts 16.5.2.6p2.  See also 16.5.5p6.  This seems
     to be already more broken than the paper's authors imagined, and more
     extensive repair is required.  I find it difficult to imagine what the
     suggested point 15 is meant to cover, since point (2) already covers
     pointer assignment which is what happens for components.  (A coindexed
     object cannot be a SOURCE= expression so intrinsic assignment is all
     we are talking about, right?)  And point (14) might be unnecessary
     once 16.5.2.2p2 is sorted out.  I'd word them all somewhat differently
     too: these are too specific, we should just have a general rule (and I
     think a general rule is easy to formulate).

  08-228r1.

     This trivial insertion surely has an incorrect cross-reference; if I
     look up vector subscripts I want to be pointed to the semantics of
     vector subscripts, not the entire syntax description (with no
     semantics) of any kind of array element or array section.  (And in PDF
     form the syntax item <section-subscript> is already hot-linked if we
     wanted that.)

  08-230r3.

     Why do we want to have a block-insert of a trivial interp request in
     the middle of the standard?  We don't need 20 lines of example code
     and 8 lines of text to beat an obvious issue to death.  Talking about
     possible processor extensions is misleading (it's just as effective
     for standard intrinsic references).  Finally, if you want to make a
     recommendation that should be in normative text.  We should not be
     making this recommendation in any case - apart from compiler test
     cases this particular feature is likely to be used exactly when the
     user WANTS it to happen.  There are MANY features in this language of
     ours that are MUCH more dangerous than this one, and for which we
     issue no recommendation.

  08-232r2.

     Obviously incomplete edits, since it introduces a syntax term
     definition which is not referenced anywhere.  And the syntax term
     <format-item-list> is referenced in at least 3 places.

  08-233r2.

     The second edit is just wrong.  It changes words that should be left
     alone, and fails to change the ones that need changing viz the "this
     subclause" in the first sentence.

  08-237r1.

     It is unacceptable to do wholesale replacement of entire long
     paragraphs when simple insertion/deletion instructions suffice; apart
     from unnecessary work for the editor it increases the risk of typos
     and of lost or incorrect markup.  Furthermore, this appears to
     exacerbate an existing contradiction in the text; the paragraph needs
     improvement, but this is not it.

3. Discussion

My comments about the unacceptability of "wholesale replacement" might seem
overly harsh, but the increased risks mentioned are not abstract ones: they
have happened already!  In any case the replacement paragraphs had obvious
problems in these cases, leading to their deferral whether the problems
were already present or not.

I have no problems at all with showing people what the result of an edit
would be, but if the actual change is to insert a relative clause to a
single sentence then that's what the instruction should be, not to replace
three sentences.

===END===