Being an IT Professional - whatever that means - I have been fascinated by the failure of the electronic vote counting machines in the Scottish parliament elections.
Sad, isn't it?
My main interest has been based around trying to uncover what went wrong. If I read the comments from DRS - the company behind the counting machines - can I discover what the real problem was? Or at least have a good guess?
Computerworld.com has the following section from a DRS spokesperson:
"The issue involved a blockage at the end of the counting process, which prevented consolidation of the data." The scanners used to count each vote -- or capture an image of unclear ballot papers for adjudication by the returning officer -- had worked without problems, she said.
The data "was secure in the system" but could not be consolidated into a format showing results that could be passed to returning officers. "That meant the results could not be announced."
The problem had been "a localized one," she said. "The majority [of counting centers] were absolutely fine."
Now, this doesn't give a great amount of detail away but I think it gives some idea of the architecture: there are scanners reading ballot papers at the front-line and these pass messages - images of the ballot paper - into a workflow. A good thing is that the data was "secure in the system," so the messages are atomically persisted. These messages then get passed to a central machine to be counted.
Ok. Nothing special. I know.
So I decided to look at the DRS website and a wonderful PDF which explains how it all works. It doesn't make things that much clearer but it does explain why there is a need for a central counting server; the VotaPad is supposed to be the end-user experience for a fully electronic voting experience. Seeing as the introduction of another piece of paper seemed to cause complete chaos, we can only be incredibly thankful that the Scottish office didn't decide to go the whole hog.
But still it is not clear what went wrong, all we hear is that:
"DRS Data & Research Services, which implemented the automated counting system, admitted it was experiencing problems with the 'consolidation' of the votes."
Further, the Eurosoc website states:
DRS executives have appeared on Scottish television news shows to reassure voters that they have been "having some known issues with the consolidation of results...The e-counting system has not crashed. The system remains secure and robust with all audit information logged. This is a temporary interruption to one small aspect of the overall process."
Now, in my book, an election system that cannot count votes is exhibiting slightly more than a 'temporary interruption.'
So, what went wrong? No idea. Apparently, there will be a full review so we should get all of the answers. Or not, as the case may be.
My guess?
I reckon that the huge number of spoiled ballots - over 100,000 - was an edge-case that DRS and the Scottish office had not tested. In this scenario, I think that larger than expected volumes of failures were recorded in the central database..... 'consolidation' in the DRS terminology. And, in the central database, MIS reports stopped working.....
Now, whenever I hear of problems in a central database that suddenly occur under load I immediately think of deadlocks, especially if reports - i.e. non transactional interactions - stop working whilst updates can continue (albeit more slowly).
What do you think? Any takers? What's your explanation?
