Before talking with Selker, I was squarely in the anti-DRE camp. After listening to him, I realize that there is another side to the story that is being systematically underreported by the technology press. Did he convince me? Well, let's say that I'm no longer convinced of the inherent correctness of the anti-DRE position.

So you can imagine how surprised I was by the next thing that Selker told me. "Of course," he said, "this country is going about election machines entirely the wrong way."

The current DRE machines, says Selker, are monstrosities. They cost ten times more than they should. Their designs are secret and their code is proprietary. And even worse, what precious few facts that have been revealed in public are deeply troubling.

A few months ago, the source code for a voting machine manufactured by Diebold was inadvertently left on a Web site. A group of researchers at Johns Hopkins downloaded the code and analyzed it. They found many software errors and poor design methodology. One of the most glaring problems had to do with encryption: although the computer used the DES algorithm to encrypt the votes, the encryption key was hard-coded into the program and unchangeable. A key that can't be changed offers little more security than using no encryption at all.

Instead of having US taxpayers spend more money on proprietary voting machines of questionable quality, Selker says that we should follow in the footsteps of Brazil, which deployed DREs in the 1990s and is currently working on the second generation of these machines.

Brazil's machines were designed in a transparent, public process by two of the country's leading research institutions. The national government then accepted bids from different companies who competed to build machines according to the open design. Everything was above-board-extremely important for a nation that has a history of election fraud.

These voting machines are simple, compact, functional, and have done a great job to bringing fair elections to the entire country. For example, each system operates on either wall current or on a set of self-contained batteries, allowing it to accept votes more than 12 hours deep in the Amazon jungle without having to be plugged in. The touch screens display not only the candidates' names but also their photographs-an important detail in a country where so many voters are illiterate. What's more, instead of costing thousands of dollars, each machine costs just hundreds.

The Brazilian machines are not perfect: they've been criticized because, like other DREs, they fundamentally cannot be audited after the fact. But security is a series of tradeoffs: the first electronic election in Brazil gave voters a printed receipt that the voters had to drop into a box after verifying it; this receipt was reportedly used for chain voting scams and the practice was discontinued in the next election.

Selker is convinced that DREs are the way of the future; many notable computer scientists continue to believe otherwise. "Election technology has not advanced to the point where it can provide us with electronic systems that are reliable enough to trust with our democracy," writes Stanford's Dill on his Web site, VerifiedVoting.org.

My feeling is that elections are in a mess throughout this country: voting machines are a problem, but so are the voter registration system, election-day intimidation, and the whole districting process. The problem with optical scan (the main technological competitor to DRE) is that unless the ballots are actually scanned when they are turned in by the voters, there is no way to prevent people from throwing away their votes by making minor clerical errors on the ballots.

Selker's argument is simple: paper is bad, and whatever problems are inherent in today's DREs can be overcome by an open design and review process.  Nobody else seems to be making this case. The U.S. DRE vendors want to sell high-priced proprietary voting machines. Meanwhile the academics want to stick with paper and all its problems.