For novel ideas about building embedded systems (both hardware and firmware), join the 40,000+ engineers who subscribe to The Embedded Muse, a free biweekly newsletter. The Muse has no hype and no vendor PR. Click here to subscribe.

By Jack Ganssle

A Close Vote?

Published 10/30/2006

On November 7 millions of Americans will go to the polls to try and persuade electronic voting machines to cast a fair vote. If the cynicism expressed by engineers who have been sending me accounts of problems with these devices in any way is mirrored in the results that day, Diebold's machines will decide the election.

That's not entirely fair, as other vendors of these machines, called Direct Recording Electronic Systems, or DREs, have equally-serious problems. In Virginia Hart InterCivic's machines, on the summary page where one acknowledges the choices made before pressing the "vote!" icon, are unable to display the candidate's full name and party affiliation ( Seems the font size is too big, so for someone with a long name like "James Webb" the unit shows only "James H. 'Jim'", the "Jim" being his nickname. No last name. No party affiliation.

It's hard to believe these things were tested when such an obvious flaw slips through.

Another company, Sequoia Voting Systems, is now owned (
by a company connected to whacko Hugo Chavez's Venezuelan government. Don't worry, though; there's no reason Chevez would want to interfere in an American election.

DREs are feature-rich and can even speak as an aid to the visually-impaired. In Yolo County, California some units (not sure which vendor) have surprised poll workers as they can be persuaded to talk only in Vietnamese (

The problem isn't confined to DREs. My state's governor is warning there may be a shortage of paper ballots (,0,1494590.story?coll=bal-home-headlines ), and in one county they've run out of envelopes to mail these to voters! The shortage is blamed at least partly on the e-voting problems in the September primary, as some officials are warning voters to cast a paper, not electronic, vote. Supplier of the paper ballots? Diebold.

The electronic voter check-in system, also provided by the same vendor, rebooted after every 43rd check-in (,0,990206.story?coll=bal-home-headlines ).

The irony is that Diebold makes both DREs and ATMs. The latter's software quality is held to an extremely high standard. No bank would tolerate errors. The company is clearly staffed with really smart developers capable of building world-class systems.

The November issue of the Communications of the ACM (not yet online) has a column that claims the feds exempt all commercial-off-the-shelf from any sort of audit when certifying a DRE. That includes Windows CE, the basis of most of these machines. Happily, we know CE is bulletproof from a security standpoint.

We've had six years since the 2000 debacle. The Help America Vote act is four years old. How long does it take to get things right?

In my opinion, the problem is that jurisdictions haven't demanded reliable and secure systems. It is possible to build an absolutely faultless DRE with proprietary code. But only if we demand it. Hold the vendors accountable. Conduct realistic tests long in advance of the election, and don't permit last-minute unvalidated patches.

That may be enough to overcome the growing distrust that may disgust and ultimately turn citizens away from their right to vote. But a better solution is to open the source code. Patent or otherwise protect it, if necessary, but subject the source to the "many eyes" scrutiny to help us trust that the elections are, well, if not fair, at least recorded accurately. Keep all the shenanigans to smoke-filled rooms as we'll never clean that mess up.

One thing we can be sure of. On November 8, in any closely-contested race the lawsuits will start to fly.