P-1788: Prospects for the new year
P-1788 members,
I wish everyone a happy and prosperous new year, as well as
many more after 2012.
Looking back, I continue to admire and respect the effort many of
us have put in, as well as the progress we have made in resolving
differences of opinion. I am optimistic that we can bring this
work to fruition soon.
Looking forward, I see several tasks. Of course, foremost
remains completion and processing of the actual standards
document. Additionally, I see two other tasks that we have
tended to overlook in the past:
1. Creation of an example implementation. I see this as
something we maintain and alter as we alter the standards
document. This implementation should be available to
all members, and possibly could be maintained with version
control, such as SVN. The process of developing such an
implementation will reveal ambiguities and weaknesses in
the current draft. The process might be made easier
by beginning with a current interval arithmetic package.
Candidates I see are (not necessarily in order of desirability):
INTLAB (Siegfried Rump's Matlab package), PROFIL / BIAS,
the BOOST interval arithmetic class, Filib / Filib++,
INTLIB / INTERVAL_ARITHMETIC (in our GlobSol package),
something from the Vienna group's implementations, others?
We should settle on one of these and commit to maintaining it
consistent with the current draft standard very soon.
(Which one we choose as the start point might depend on who
volunteers their time to maintain it.) I need comment,
volunteers, and possibly a motion (if there is no
unanimous consent).
2. Creation of a test suite to certify that a package complies
with the standard. Again, we should start this soon, even though
the standard is not completed yet. The test suite can be maintained in
a manner similar to the example implementation, in collaboration
with maintainers of the example implementation. I also need
comment and volunteers for the test suite.
In another matter, people have suggested that I work out a
detailed timetable (such as a PERT chart) for the remaining
tasks. I have found doing so exceptionally difficult.
However, I will have a better idea once we
have prototypes for the example implementation and test suite.
After that, much of what remains would be formal processing
of the document, as well as clearly identified remaining
issues; times and dependencies could be estimated for those items.
Best regards,
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
---------------------------------------------------------------