@mattblaze hand marked scantron is the only realistic solution. Allows quick results and a verifiable paper ballot for recounts.

Any *private* non-public source code counting our votes is absolutely undemocratic.

@apenguininspace @mattblaze that's the problem. We don't know.

A paper trail is better than nothing but there's just precious little benefit to having a software program do the voting part.

If they open source their code so it's reviewable, that would be ok...but still overkill for what's needed, simple scan n read.

Next up is the blatant security failure of these *private* machines. Sure they've been improved, but why incorporate the risk if you don't have to?

https://m.slashdot.org/story/298229

@pixelpusher220 @apenguininspace Here’s the problem: today’s elections in the US are incredibly complex compared with the rest of the world. We vote on a ton of different races, and it’s simply not feasible in most of the US to tally ballots without automation. But automation is inherently unreliable and insecure.

Risk limiting audits are an efficient way to get the benefits of automation while also assuring against errors or compromises of the tally system.

@mattblaze @apenguininspace yes, scantron is automation.

My issue is unaudited, untested machines running unaudited code. The examples of bad (or basically no) security are legion as any private actor will only do the minimum.

I still posit that the software voting machines provide no significant benefit, let alone when weighed against the increased risks.

Edit: *publicly audited

@mattblaze @apenguininspace

Agreed it adds layers of risk...that are almost entirely avoided by scantron counting.

Risk limiting audits can be done regardless of the counting method.

What measurable increase in reporting does software voting provide?

@mattblaze @apenguininspace @pixelpusher220 I was really bothered when I took a look at the ballots printed by our Dominion ballot printers here in Santa Cruz.

It had both human (and probably machine readable) vote choices, but also had a QR code that was damnably difficult to read using standard QR reading software and all one got was a binary blob that had to be interpreted via a Dominion proprietary decoding table. And to add butter to the popcorn the QR code was used by the ballot readers as the primary representation of voter choices.

It seems to me that a ballot to be scanned should have exactly one - not multiple - representation of the voter's choices, and that representation should be comprehensible by people using normal human senses.