NA Digest Monday, December 26, 1994 Volume 94 : Issue 52

Today's Editor:

Cleve Moler

The MathWorks, Inc.

moler@mathworks.com

Today's Topics:
• Birthday Puzzle
• Charta of Free Electronic Access to Publications
• Pentium Hypertext and the Sofware Taxi
• Prelude to Pentium
• LIPSOL Beta-2 Release
• New Mail Address for Los Alamos PPNS Group
• New Book, Practical Bifurcation and Stability Analysis
• New Book, Stochastic Programming
• South African N.A. Symposium
• Conference on Parallel Algorithms
• Position at University of the Federal Armed Forces, Munich
• Position at Illinois Institute of Technology
• Contents, SIAM Applied Mathematics
• Contents, SIAM Optimization
• Contents, Journal of Approximation Theory

Submissions for NA Digest:

Information about NA-NET:

-------------------------------------------------------

From: Gene Golub <golub@sccm.Stanford.EDU>
Date: Mon, 19 Dec 94 8:35:20 PST
Subject: Birthday Puzzle

What TWO famous numerical analysts will be celebrating their 60th birthday
on Dec 26, 1994? Hint: Their birthplaces are antipodes.

I'm sure they'll be happy to hear from you.

Gene


------------------------------

From: R. Jeltsch <jeltsch@sam.math.ethz.ch>
Date: Mon, 19 Dec 94 15:32:35 +0100
Subject: Charta of Free Electronic Access to Publications

Charta of free electronic access to publications:

The EMS (European Mathematical Society) has set up
a committee to draft a Charta of free electronic
access to publications. I am on this committee and
would like to get some input from the NA-NET community.
Please respond either to me or to the electronic
internet address
ems-e-jour-committee@esi.ac.at
distributes every message it receives to all members
of this committee, which consists now of
bayer@grenet.fr (Eva Bayer)
brunovsk@dcs.fmph.uniba.sk (Pavol Brunovski)
chlebik@dcs.fmph.uniba.sk (Miro Chlebik)
cea@unice.fr (Jean Cea)
groetschel@ZIB-berlin.de (Groetschel)
guil@fourier.grenet.fr (Laurent Guillope)
jeltsch@sam.math.ethz.ch (R. Jeltsch)
amo@research.att.com (Odlyzko)
uimrozek@plkrcy11.bitnet (Marian Mrozek)
michor@esi.ac.at (Peter Michor, chair)

The first draft of such a charta has been proposed
Peter Michor and he has presented it at the
Conference on the Future of Mathematical Communications
electronic publishing MSRI, Berkeley, Nov. 30 -- Dec. 3.

First draft (shortened for Na-net)

The role of a printed copy of a journal nowadays is sitting in a
library waiting for an article out of it to be photocopied. A
conservative estimate is that on the average an article is
photocopied 50 times, all over the world, and read 25 times. To
typeset, print, and distribute it is a waste of human and natural
resources, the same service could be done via computer quicker
and more widespread.

There is an oncoming revolution on the use of electronic media for
scientific publication. New electronic journals appear, they are
FREELY accessible up to now, some existing journals open their files
for FREE electronic access, and some only to subscribers. There is
the possibility to lower drastically the overall cost of mathematical
publications, and the danger that we will have all the traditional
journals, new electronic ones, and that after some time the
electronic journals also start to charge for their access. If there
is no easy access to the mathematical literature, a lot of work will
be duplicated in the future.

The European Mathematical Society feels responsible for the health
and accessibility of the mathematical literature.

The rationale behind this charta is the following. We should
distinguish carefully between the final electronic version of a paper
into which besides the work of writing and refereeing only marginal
costs have been invested, and the printed and distributed version,
which makes use of a somewhat industrial process. The costs of the
former, since they are small, should be carried locally, by academic
institutions or libraries. Important libraries could adopt existing
journals and take over their costs up to the electronic product. Free
personal access to the electronic product should be ensured. The
costs of the latter industrial process should be borne by the
consumer (library) who prefers to have at hand a nicely printed
version. There will be not so few of them, perhaps.

CHARTA OF FREE ELECTRONIC ACCESS TO PUBLICATIONS:

1. The right of access to the electronic file of a paper which was
prepared in that way by the author or his institution lies with the
author(s).

2. If the author submits this paper to a preprint server or a
refereeing process he automatically acknowledges the right of free
electronic access to this paper to the (mathematical) public.

3. If a journal prints a refereed article the right to distribute and
sell this paper lies completely with the publisher of the journal.
This is called the RIGHT TO DISTRIBUTE. But the right of free
electronic access to the paper remains at the public.

4. Everybody who uses the right of electronic access to a paper is
entitled to print freely one copy of this for personal or restricted
use. He is not entitled to sell copies of this paper at profit. This
is called PERSONAL USE.

5. Libraries are entitled and asked to keep permanently accessible
files of electronically accessible journals, in order to ensure their
perpetuity, besides or without printed copies.

We ask authors, readers, institutions and libraries, profit and
non-profit oriented publishers to adopt the rules of this charta, for
the benefit of the science of Mathematics, and thus also for the
benefit of those commercial organizations serving it.


------------------------------

From: John C. Nash <jcnash@aix1.uottawa.ca>
Date: Mon, 19 Dec 1994 11:51:58 -0500 (EST)
Subject: Pentium Hypertext and the Sofware Taxi

Mainly for the benefit of non-NA colleagues and for some local
computer store operators, I edited and collected commentary and test
codes for the Pentium FDIV bug. These have been built into a
hypertext form so the comments can be viewed easily and a 1-key
command will execute Mathisen's test.

The hypertext viewer is my own Software Taxi which NA Digest readers
may also find interesting as a way to present "live" demonstrations
of software along with commentary. Simple text files are used for
the hypertexts and control files. Freeware.

The Pentium hypertext and the Software Taxi are available by
anonymous FTP from either

MacNash.admin.uottawa.ca

(look in the "binaries" subdirectory)

or ftp.synapse.net in the info/nashinfo/ subdirectory.

The files in the LHA archive PENTIUM0.LZH contain information,
comments and test programs about the Intel Pentium FDIV bug. They
are designed to be used in conjunction with the Software Taxi
hypertext viewer in SWTAXI0.LZH.

You should unpack the files from both archives into the same
directory (or diskette -- 1.2 MB or 1.4MB format). To view/run the
files, type

ST0 PENTIUM0 [Enter]

To run the Software Taxi general introduction, simply type

ST0 [Enter]

and accept defaults. The Software Taxi is only 23K and designed for
student use on low-power PCs; it is fast enough to run from a
diskette.


John C. Nash, Professor of Management, Faculty of Administration
136 Jean-Jacques Lussier Private, University of Ottawa, Ottawa
Ontario, K1N 6N5 Canada jcnash@aix1.uottawa.ca fax 613 564 6518


------------------------------

From: Garry Tee <tee@mat.auckland.ac.nz>
Date: Thu, 22 Dec 1994 17:23:12 +1200
Subject: Prelude to Pentium

In 1960, when Leslie Fox was Director of Oxford University
Computing Laboratory, a Sirius computer (of the 2nd generation) was
installed there. Fox had gained his computing experience on Turing's Pilot
ACE (built in 1950 by J. H. Wilkinson) and on its commercial version DEUCE;
and hence he was accustomed to programming in which every bit was correct,
including fixed-point arithmetic in which round-off was in the range -1/2
to +1/2 bit in the least significant place.
His testing of Sirius had revealed persistent errors in the arithmetic
performed; and eventually he was horrified to discover that the arithmetic
hardware performed fixed-point division with roundoff between +1 and +2
bits! Leslie Fox erupted with scandalized outrage at that inexcusable
blunder by the manufacturers of Sirius!

Garry J. Tee, University of Auckland.


------------------------------

From: Yin Zhang <yzhang@math.umbc.edu>
Date: Wed, 21 Dec 1994 13:54:26 -0400
Subject: LIPSOL Beta-2 Release

LIPSOL Beta-2 Release Announcement

I am pleased to announce the Beta-2 release of LIPSOL -- Linear
programming Interior-Point SOLvers. LIPSOL is a package that uses
Matlab's sparse-matrix data structure and MEX utilities to achieve
both programming simplicity and computational efficiency.

--- What's New? ---

Release Beta-2 is a major step-up from the first Beta release (now
called Beta-1) with important enhancements and bug-fixings. The
highlights are:

1. Beta-2 now solves 94 Netlib LP problems out of 95 (fit2p, still
untested, is too costly without a dense-column handling strategy);
while Beta-1 could only solve 80 problems. A single set of default
parameter values is used to solve 93 problems (except for dfl001).

2. Beta-2 is much faster than Beta-1 for large problems. Here are
some statistics on an SGI machine (R4400/150Mhz processor) for 3 of
the largest problems:

                                           CPU-seconds

          Name      rows   cols   iter   Beta-2   Beta-1

       ----------------------------------------------------

         pilot87    2031    4883    40     569     2315

        maros-r7    3137    9408    14     508   untested

        stocfor3   16676   15695    38     206   untested

       ----------------------------------------------------