Paper Vote Canada

Delays, equipment breakdowns and erroneous results marred the IT systems used to handle the recent municipal election of Montreal and several cities in Quebec.

The voting box IT infrastructure was supplied by PG Elections, an affiliate of PG Mensys, which has helped to run several elections since 1999. PG Elections deployed some 1,400 electronic ballot boxes and voting terminals in 604 sites, providing an electronic voting system in a majority of municipalities in the province of Québec (a number of others cities also voted electronically using systems deployed by another service provider). Some of the machines failed to count properly and a mayoral contender in Montréal called for a legal recount in several districts after noting a number of problems in the voting process.

A voting terminal is a completely electronic device that registers votes using a touchscreen display, but the systems that failed during these municipal elections across the province were electronic ballot boxes used to scan paper ballots, which are then digitized and compiled. Some 900 of those voting terminals were deployed, of which 450 were rented from an American supplier. Thomas Gagnon, president of PG Elections, refused to name the partner who supplied the electronic voting systems.

“Everything worked well in the testing, but when we deployed them in the field, there were a series of important problems: if someone voted too quickly, they broke down, but if you voted slowly enough they worked well,” said. “We had what you could call lemons.”

Poor Internet connections also induced delays in the transmission of the electoral results to the political parties and to the media, who had to resort to more manual methods by telephones. “That was a major irritant,” Gagnon admitted.

Gagnon said about 45,000 ballots were counted twice, but that they were corrected before the end results were announced.

PG Elections attributed the catastrophe to an excess of optimism and for providing a team of only 300 technicians to handle all of Quebec.

Emphasis mine. "An excess of optimism"? How about a shortage of caution.
from IT Business / Computing Canada - Montreal's electronic voting debacle: What went wrong - November 22, 2005

Translated (by IT Business) from the original French on their sister site
Direction Informatique - Histoire d'une catastrophe informatique - November 11, 2005

11/11/2005 - Au-delà de la compétence de la firme impliquée, PG Elections, les ratés informatiques des dernières élections municipales au Québec ont fourni la preuve que la loi de Murphy existe

Il y a peu de chances que Thomas Gagnon, président et chef de la direction de PG Mensys Systèmes d’Information, garde un bon souvenir du 6 novembre 2005, journée d’élections municipales généralisée pour la première fois à l’ensemble des villes du Québec, ou presque. Une journée qui devait selon toute probabilité se dérouler sans accrocs, du moins d’un point de vue informatique, l’infrastructure de vote électronique fournie par PG Elections, une filiale de PG Mensys, ayant été testée et utilisée à plusieurs reprises depuis 1999.

Oh baloney. Murphy's Law? "selon toute probabilité se dérouler sans accrocs"?

You know how you succeed with stuff? You PLAN. You MANAGE THE RISKS.

Or, apparently if you're PG Elections, you think "Hey, what could possibly go wrong? To the Vote Machine!"

And then you say "sorry we screwed up your elections, hey, guess what, turns out we didn't know what we were doing, we were optimistic without reason!"

You know what machines do?
They break.
They fail.
They malfunction.
They're misprogrammed.
The network goes down.
They overheat.
etc. etc. etc.

You know what a paper ballot does?
Err, it just sits there, working fine.
Other than setting it on fire, it's pretty hard for a paper ballot to malfunction.

Which one is better suited to a task that must work perfectly in a single, specific day?

And I don't want to rag on PG specifically.
Voting using machines is very hard to do well, even if you're not "optimistic".
It's a fundamentally flawed concept.

Voting happens rarely, the entire complex system has to work perfectly for millions of people in numerous locations... this is not a set of requirements that computer technology and business planning processes are well-suited to solve.

Businesses are suited to making lots of money by doing frequently repeatable transactional things as "efficiently" (e.g. as cheaply and poorly as they can get away with). That is not a great model to apply to elections.

And even if you had teams of computer scientists working very hard, it wouldn't work well.

The fact that the actual voting machines, systems and processes are generally hacked together with the minimal possible investment of time, thought, risk assessment and computer security expertise just makes things all the worse.