P-1788: Rough plan and request for volunteers
P-1788 working group:
I have reviewed my notes and key documents, in an attempt to pin down
use of the remaining time allotted to the P-1788 working group. (The
term ends December 2012, that is, after 64 more weeks, but can be extended,
with justification.) Doing so, there are "fundamental" or
"guiding" issues, that are more-or-less independent of each other, and
other issues that build on these, or are more complex. (What I outline
below is essentially a synthesis of notes Nathalie has provided me,
as well as Arnold, John, and Nate's work and comment.) After that, we need
to vote on actual standards wording. Prior to resolving the primary and
derived position issues, I do not see a way to precisely predict when
we might finish the actual standards document wording. (Perhaps, could
I use an interval PERT chart?)
In any case, here are the items, each of which seem to require
one or more motions.
Primary questions:
------- ---------
1. Should union and intersection be excluded from expressions in which
decorations are processed?
2. Should the 1788 standard specify levels 3 and 4 at all?
That is, should the internal representation be opaque?
(Issue: even using 754 data types and inf-sup, different internal
representations are possible)
3. Should the 1788 standard specify a data type for which the
results are reproducible across platforms?
(In accordance with our previous discussions on the subject,
we should word such a motion in a reasonable way, since we
have identified multiprocessing situations in which such
reproducibility may not be possible.)
4. Should the dot product mandated in Motion 9 (which passed) be
faithful or accurate?
("Accurate" means the nearest floating point number to the
exact result, whereas "faithful" means, I think, to within
one or two ULPs of the exact result -- correct me if wrong.)
I think "exact" implies "accurate," but it was requested that this issue
be revisited.
Note: Revol, Lefevre, and Rump have already volunteered to produce a
position paper and motion on this, with me as acting
chair, so Nathalie can actually participate in the
motion.
Compound and derived questions:
-------- --- ------- ---------
1. Specification of a specific set of decorations and corresponding system,
upon which there is a consensus (or near consensus).
2. Interval constructors
(This is tied to the decorations system, such as use of Arnold's
"domain" function to initialize the decoration. However, there
are other, specific issues, such as handling Interval(Inf,Inf), where,
as decided in Motion 3, [Inf,Inf] is not in the set of intervals. I
am confident such things can be agreed upon in a simple, elegant
way.)
3. Compressed intervals (This is a separate issue associated
with decorations, where a more efficient version of decorations
is proposed in contexts not requiring the full system. Nate has
privately expressed a promising idea of how this can be
specified, without discarding too much information.)
4. ???
Additional, more routine questions:
---------- ---- ------- ---------
1. Specification of the set of supported / recommended standard functions.
2. Work on a test suite to verify that a system complies.
(not a motion, but a lot of work that cannot be completed until after
the other issues are resolved).
3. ??
Things I missed:
------ - ------
1. ?????
===============
I need a "champion" to write up and move each of these issues. My perception
is that the primary questions should be resolved first, but may be resolved
in parallel. My advice is to write up the primary motions in a short, simple,
and clear way, although supporting arguments (clearly identified as such)
can be submitted separately. (For example, I envision the motion corresponding
to primary question 1 as consisting of a single sentence.) In my opinion,
we will receive clearer guidance and process items more efficiently if
we separate out more controversial items and process them separately, first,
in simple motions.
Thus, please volunteer and commit to an issue. When you do so, please
send your intentions to me, so I know what to expect.
At a recent meeting at Schloss Dagstuhl, it was requested that I plan the
time to the end of our initial term. Although I still remain optimistic
that we can complete within [1,4] years, there seems to be too much
uncertainty in the working group's actions for me to make solid
prediction. The lower end of this range will happen if the primary
questions are resolved quickly (say within the next 2 months), if the
derived questions are resolved within the next two months, and if voting
on the standard document wording proceeds in an orderly way, without
too much revisiting of issues resolved in position papers. Once remaining
position issues are resolved, John and I can work out a more accurate
(but still tentative) schedule for processing the actual standards
document wording.
To facilitate the process, I request that working group members read and
review each issue carefully, that they put appropriate subject lines in
their emails, that they remain on-topic (or on-one-of-the-topics when
multiple issues are being simultaneously processed) unless overlooked
issues need to be FORMALLY introduced, and that appropriate thought
be put into composing postings.
Sincerely,
Baker
--
---------------------------------------------------------------
Ralph Baker Kearfott, rbk@xxxxxxxxxxxxx (337) 482-5346 (fax)
(337) 482-5270 (work) (337) 993-1827 (home)
URL: http://interval.louisiana.edu/kearfott.html
Department of Mathematics, University of Louisiana at Lafayette
(Room 217 Maxim D. Doucet Hall, 1403 Johnston Street)
Box 4-1010, Lafayette, LA 70504-1010, USA
---------------------------------------------------------------