Most computer-savvy people I know would never entrust something as important as picking the next president or governor to gadgets as unreliable as computers. Especially with no paper backup.
Still, I was amused when our governor denounced Maryland's $55 million, Diebold electronic voting system as too flaky and unreliable to count the votes in his 2006 re-election campaign.
Cynics might call Bob Ehrlich's last-minute conversion a political ploy. After all, he's trailing in the polls and needs a convenient scapegoat for losing the election - or for contesting the results.
Cynics might also note the departure of Diebold's longtime chairman, Wally O'Dell, in December.
O'Dell is an outspoken Republican fundraiser with close ties to President Bush. While he was running Diebold, Republican pols were reluctant to attack it.
With O'Dell gone, some say it's open season on the company.
But I'm not that cynical. I'm just glad that another prominent voice has joined the campaign against a voting system whose very design leaves it open to charges of tampering and mismanagement.
The question is, what do we do now?
One possibility is legislation that would add a so-called "paper trail" of printed ballot receipts to Diebold's all-electronic, touchscreen voting machines.
But a task force at the University of Maryland, Baltimore County told the elections board that none of the add-ons currently available for Diebold machines was guaranteed to be reliable - and certainly not in time for the 2006 election.
But here's the real reason why Diebold's system and others of its ilk aren't safe and never will be: No one can see inside them. No one can watch them count votes, literally or figuratively.
Remember the old days of paper ballots? Election judges (Republican and Democrat, side by side) looked at each one and marked down the results.
Both parties had observers. If a ballot counter tried to record a Smith vote in Jones' column, someone would be there to catch it. In case of a dispute, the ballots were available for a recount.
The voting system used by Baltimore County and many other jurisdictions before Diebold was an electronic variant of the original paper system. It used paper ballots but scanned them electronically to produce instant tallies on election night.
If a dispute arose or a scanner broke down, the originals were available to be rescanned or counted by hand. A study by Yale University - one of the first serious efforts to rank different systems of counting votes - found that this hybrid system produced the most accurate results.
The problem is that scanners were too cheap, too simple and too readily available when Congress dumped billions of federal dollars on state governments to upgrade systems after the 2000 presidential election fiasco.
There was enormous pressure to spend that money on something more expensive and allegedly more advanced - Direct Recording Electronic systems, or DREs.
Besides the gee-whiz factor, all-electronic systems fulfilled every election administrator's dream - they eliminated paper ballots. Administrators hate paper ballots. They're expensive to print and store, and in boxes they're too heavy for elderly voting judges to handle.
But for the public, they're also the best guarantee of a fair, honest and accurate vote count.
I don't have space here to go into the security issues and controversies that all-electronic voting machines have generated in other states. But the ultimate problem is that all counting in DRE systems is done by software - a set of instructions that tell the system how to record every vote cast. Diebold uses "proprietary" software, a buzzword meaning "secret." The source code belongs to Diebold. You and I can't examine it.
That software could easily record every 99th Democratic vote for a Republican, or vice versa. It would be nudged just enough to skew a close race without detection. Remember that Maryland's 1994 gubernatorial election was decided by less than half a percentage point.
Computer glitches like this can be the result of tampering, a hardware malfunction or just plain bad programming. But the way things stand, you and I will never know.
In fact, the revolt against Diebold and other electronic voting vendors occurred because employees left Diebold's source code on an unsecured Web site, where a bunch of security experts from the Johns Hopkins University got hold of it and exposed its flaws.
Without access to that code, all we have is a black box. There is no original voting record - just a handful of electronic 1's and 0's that are supposed to represent the buttons we pushed. As things stand, there's also no paper trail of any kind - a longtime issue with critics of electronic voting.
Here's another issue: With every jurisdiction in Maryland now using the same system, all it takes to ruin an entire statewide election is a single glitch in a single line of that secret code.
In the systems business, this is known as a computer monoculture. It's a term borrowed from agriculture to describe a large area planted in a single crop - and hence vulnerable to devastating damage from a single source. Maryland is completely planted with Diebold's electronic cotton - all it needs for disaster is one electronic boll weevil.
To all of this criticism, Linda H. Lamone, the state election administrator has had one response: "Trust us."
Well, I don't and you shouldn't. Elections aren't based on trust. They're based on verifiable results. You can't throw technology at a problem and throw common sense out the window. There's no way to fix this system. I don't care how much we've spent on it.
So let's bite the bullet and use Diebold, very gingerly in 2006, because we'll only make a bigger mess if we try to rush changes before the election. But let's start shopping right now for a better system. There are updated, reliable paper scanning systems available, and we'll soon see new generations of DREs that provide paper backups and better verification.
This won't be cheap, but Maryland has a billion-dollar surplus this year, and we can use a tiny fraction of it to buy a good election system - one that uses open source code that everyone can see (and test), plus a workable paper backup.
It's a small price for a generation of elections we can believe in.