Over the years I've always been a fan of shareware and freeware. People working on their own time, often at home, or just as a hobby have certainly come up with some ingenious and useful programs. In fact, there are some programs of this class I use almost daily for such tasks as file compression or graphics manipulation.

On the DSP side, you can find plenty of interesting shareware, as well. My list of links elsewhere on this DSP page points out a few of the ones I've located. I find especially fascinating some of the Matlab knockoffs. I'll admit they don't have the completeness, polish and support of Matlab, but they cost thousands less and perform many of the same functions. Some detractors correctly point out that you don't always know exactly how these packages implement a given algorithm, and even if the source code is available, who wants to take the time to step through every line to see how it functions? That's one great big advantage Matlab does bring; you can have inherent trust in the quality and reliability of the underlying algorithms.

Does that mean there's no room for freeware/shareware in the Matlab space? Not by any means. One such program caught my eye recently and I'd like to bring it to your attention. It addresses the fact that even though PCs are getting bigger and faster, they still don't always keep up with the complexity of the problems we throw at them. I think the DSP area is especially prone to this effect. As a result, a grad student at Purdue University (Tom Kraus) has put together some software that parcels bits of a large simulation onto different machines to increase the speed of execution. The result is a set of packages called CommSim and the Multi Toolbox, which he distributes at no charge. Note that the set is intended initially for users who own multiple Matlab licenses and have several Unix machines at their disposal. One reason for this restriction is that the tools require the rsh command, which must be enabled to work without a password.

Multi consists of a set of tools for starting Matlab processes on remote machines, managing a list of processes, executing Matlab statements in a nonblocking fashion on those remote machines as well as passing Matlab arrays between the master process and the children. One part of the toolset is CommSim, a graphical interface implemented as an M-file that allows you to call a Matlab script or function many times with different parameters, all on different machines. In short, a parent Matlab process handles the bookkeeping and scheduling of a simulation. Multiple child Matlab processes run in the background, and the parent process uses realtime scheduling to decide which pieces of the simulation to distribute to the children.

I think a few of you might enjoy taking a look at this package. And moving from this specific example to the more general case, I think that the entire user community is taking a fresh look at nontraditional software sources. Perhaps the single biggest event in that regard is the wild success of Linux. Here again documentation isn't everything you'd like it to be, and support can become an issue, especially if you start playing around with the source code. Nonetheless, several companies have taken this software and packaged it in a way that makes it easier to use, and in the process they've become quite successful and made Linux almost mainstream in some market segments. Further, Linux is now spinning off some related products that are following the same distribution model. For instance, you can now find several sources for software modules that make Linux a realtime OS, and this alternative to full-blown RTOSs, embedded kernels or NT add-ons is attracting interest from some major companies and organizations.

Clearly, nontraditional software is overcoming its bad rap from the past. Some very bright people are applying their talents and sharing their results in this way. We should all keep our eyes alert and our minds open to the possibilities.

		Click here to get your listing up.