Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

P1788: M0006.04_Level_2_Multi-format PASSED



P1788,

M0006.04 Level_2_Multi-format is PASSED
   Yes: 48, No: 4, of 71 registered Voting Members
   Quorum is half the voting members
   Passage is by half those voting


PLEASE VOTE ON MOTION 7.  Current tally:
   Yes: 13, No: 14, of 71 voting members (ends 21 Sept.)

George Corliss, P1788 Voting Tabulator

The motion text is as follows:

===============================================================
P1788 members (from John Pryce)

As requested I circulate herewith version 4 (hopefully final) of
Position Paper 13, "TEXT AND RATIONALE FOR MOTION 6: MULTI-FORMAT
SUPPORT", on which the vote will be conducted.

Also as requested by the Chair, here is a brief change-history.

2009 April 21. Original version of motion 6 circulated in plain text
form.

June 9. Following extended criticism and discussion, version 2 of
plain text has a clearer statement of the motion, new "Overall aims"
section, improved Definitions, and introduces the SRSF, SRMF, MRMF
concepts.

June 29. Typeset version produced as a Position Paper, mostly similar
to previous plain text.

July 14. Typeset version, revised, circulated as Position Paper PP013
and motion 6 formally re-submitted. Revisions with help from Arnold
Neumaier, Ulrich Kulisch and others improve notation and make it
consistent with Motion 5. Statement of motion simplified and made
more precise.

July 14 - August 11. Meaning of "interval mapping", "interval
function" clarified. After comment by Vincent Lefevre, abstract/
concrete "floating-point format" renamed "number format" (so af-
format, cf-format become an-format, cn-format). Other minor
corrections. Vincent also proposes support for MRMF in "valid" mode
be included in motion.

Today Aug 11. In this final version I have modified what is said
about MRMF support in the rationale, but have not included Vincent's
proposed words in the motion. Instead, the Chair has suggested
Vincent proposes a separate motion covering MRMF support.

Best wishes

John Pryce

================================================================


For the record, NO votes and comments.  Technical Editors may consider these
when framing subsequent motions:

From: Maarten van Emden <vanemden@xxxxxxxxxx>
To: <stds-1788@xxxxxxxxxxxxxxxxx>
Conversation: Motion 6

My vote is NO.

Rationale: Motion 6 addresses an issue at level(s) below level
1 of the document accepted as per Motion 2, while there are
unresolved issues at level 1. One of these issues is the definition
of relational operations (at level 1, of course).


From: Frédéric Goualard <Frederic.Goualard@xxxxxxxxxxxxxx>
To: <stds-1788@xxxxxxxxxxxxxxxxx>
Conversation: Motion P1788/M0006.04_Level_2_Multi-format NO

My vote is NO because the associated position paper definitely
assumes---even if peripherally---the implicit (e.g., [-oo, -oo] and
[+oo, +oo] have no meaning, which should make them NaIs, whatever that
object is) or explicit existence of NaIs, a concept I am not in favour of.

FG.
- --
Frédéric Goualard                                 LINA - UMR CNRS 6241
Tel.: +33 2 51 12 58 38    Univ. of Nantes - Ecole des Mines de Nantes
Fax.: +33 2 51 12 58 12            2, rue de la Houssinière - BP 92208
http://goualard.frederic.free.fr/               F-44322 NANTES CEDEX 3



From: Dominique Lohez <dominique.lohez@xxxxxxx>
To: George Corliss <George.Corliss@xxxxxxxxxxxxx>
Cc: stds-1788 <stds-1788@xxxxxxxxxxxxxxxxx>
Conversation: [IEEE P1788] P1788/M0006.04_Level_2_Multi-format NO

My vote is NO

Dominique LOHEZ

I would vote YES if the following changes were done

        1) The subsection 3.5.2 is removed
         2)  In subsection   3.5.4    the statement
                             1. Here, f is a mapping, not an expression.
               should be completed by explaining what is an expression
and why the difference  is important   in interval arithmetic.

         3) In subsection 3.6.1 the  reference   to R* should be avoided.
              This can be achieved by  considering a lattice with a
finite number of elements and a linear order relation.
               Such a lattice has a minimum which can be labeled  -Infinity
                Such a lattice has a maximum which can be labeled  +Infinity
                 the other elements  are an embedment  of the
appropriate finite subset of R
                 
Rationale
     1) Since the author's intent of is that  the text is  used as it in
the standard. All the wordings and statements  must be checked carefully.
     2) Concerning the change 1
                -Either the decision has been taken when the Motion  was
approved. And the meaning should have been explained as an annex of this
motion.
                  ( It is my own opinion)

                 - Either  there a new  choice concerning  (-Infinity ,
- Infnity)  ...   and this choice should not be dealt  as a side effect
of a motion devoted to
                    a completely different subject.
    3) The standard must be readable to new comers in  Interval
arithmetic (Change 2)
    4) I cannot prevent myself from thinking  when I encounter a
reference to R* to a surreptitious  come back  of  what was rejected in
Motion 3 .
         And a simple alternative is available here.

            --
Dr Dominique LOHEZ
ISEN
41, Bd Vauban
F59046 LILLE
France

Phone : +33 (0)3 20 30 40 71
Email: Dominique.Lohez@xxxxxxx



From: G. William (Bill) Walster <bill@xxxxxxxxxxx>
Reply-To: <bill@xxxxxxxxxxx>
To: <stds-1788@xxxxxxxxxxxxxxxxx>
Conversation: Motion 6

I vote NO on Motion 6.

There are good reasons for interval data types to be opaque.  This means
that the internal representation of an interval should only be
accessible through intrinsic functions supplied with the given language
support for interval data types.

The advantage is that hardware manufacturers are thereby free to choose
an internal representation that best suits their machine architecture.
It also means that any "standard" internal representation cannot act as
an roadblock to mathematical and implementation innovations that produce
superior quality results in terms of speed and width.

Ideally, this leaves only one thing to be done in an interval standard:
Define the set of values that any compliant interval implementation must
contain.  How such a standard is implemented can then be totally left up
to the imagination and creativity of different implementers.

As far as I can tell, this most important definition does not yet exist.