Dr. Dobb's Journal November 2003
The National Medal of Technology is awarded to "individuals, teams, and/or companies/divisions...who have made lasting contributions to America's competitiveness, standard of living, and quality of life through technological innovation." First awarded in 1985, the medal has in the past gone to individuals such as Dean Kamen, institutional directors such as Arun Netravali of Bell Labs, and corporations like IBM and Dow Chemical.
Nominations for the 2004 medal may be made at http://www.technology.gov/Medal/ until October 30. An evaluation committee appointed by the U.S. Secretary of Commerce will review the nominations and suggest a candidate to the Secretary, who will pass on his recommendation to the President. This year's evaluation committee consists of eight members, including Steve Wozniak as chair.
The DotGNU Project has announced a "collaborative coding competition" (http://www.gnu.org/projects/dotgnu/competition .html). Participants will work together to create a Free Software implementation of the System.Windows.Forms part of the C# class libraries. After four months, or when the project is complete, five of the most active contributors will receive cash prizes for their participation. Participants will be judged not only on the code they contribute, but also on their cooperation with other members of the team and their willingness to help newcomers. The highest prize will be $2000, the lowest $200.
The DotGNU project is, with the Mono initiative (http://www.go-mono.org/), one of two groups working to build an open-source architecture compatible with Microsoft's .NET platform. The System.Windows.Forms functionality is required by the DotGNU Portable.NET branch of the project, which is working on Free Software tools for executing Common Language Infrastructure applications.
Winners will be chosen by a three-member jury, advised by DotGNU Portable.NET chief architect Rhys Weatherley. To be eligible for the prizes, participants have at least one code contribution accepted into the pnetlib System.Windows.Forms codebase.
The Apache Software Foundation has launched a new project to build a new open-source J2EE server, with the goal of achieving certification from Sun. The Geronimo project (http://incubator.apache.org/projects/geronimo.html), as it is dubbed, incorporates a group of developers formerly associated with The JBoss Group. JBoss is one of a couple existing open-source J2EE servers; however, neither are Sun certified, and both are released under the LGPL rather than Apache's less restrictive BSD-style license. And with Geronimo, Apache will tie existing projects that implement pieces of J2EE functionality (such as Tomcat) into a full-fledged J2EE stack.
The Geronimo developers are confident about obtaining J2EE certification from Sun, pointing out that Apache has already worked with Sun to get Java certification for the Tomcat module. While the JBoss code cannot be reused in Geronimo due to the terms of the LGPL, the developers expect to draw from many existing projects within the Apache umbrella and a few outside codebases such as OpenJMS and the Tyrex transaction manager.
A pair of researchers say they have identified a flaw in the LFSR (linear-feedback shift register) class of pseudorandom-number generators, which may explain the failure of some "random walk" simulations. Stephan Mertens of the Abdus Salam International Centre for Theoretical Physics and Heiko Bauke of Otto von Guericke University published a paper in the journal Condensed Matter that shows a bias in random-number generators based on arithmetic modulo 2 (http://xxx.arxiv.org/pdf/cond-mat/0307138). These generators are sometimes used in simulations to give the effect of a coin toss—for example, a 1 could be "heads" and a 0 "tails." When tested, however, it turns out that this virtual coin flip is actually more likely to produce heads than tails.
"The problems arise because of the special role of the zero in the algebra of finite fields," say Mertens and Bauke. "Random-number generators that avoid (or ignore) the zeroes...are even better than real, physical coins." In fact, they suggest that with this information in mind, LFSR random-number generators should be considered more reliable than ever: "Better to know and to control the deficiencies of a random-number generator than to rely on fancy methods, which are basically justified by empirical observations."