I've seen XKCD#2030 go around here too, and people mostly think positively of it. I'm sorry, but you're wrong.
Thankfully, someone a lot smarter than me already wrote a post about why: https://securityboulevard.com/2018/08/that-xkcd-on-voting-machine-software-is-wrong/
TL;DR: Elevators and Air planes are designed to guard against accidents. 2030 compares that to vulnerability to active attacks. Airplanes and Elevators are even worse in those situations than most software. It's not a good comparison.
@algernon Computer programs contain errors, always. An error in a lift program can damage less people than an error in an airplane's piloting program. How many people can damage an error in a voting machine? Sorry for my English that comes from a machine translation program, which contains (surely) some errors. 🙂
@pepsy The thing is, it's not that voting machines are good or bad, or dangerous or not. The problem with XKCD#2030 is that it makes an unfair comparison. Like, if you compared the severity of a kitchen accident (smart toaster burns your toast) vs an active terrorist attack (terrorists take over a plane's computer), to prove that software development is terrible. Both are software "errors", but one's an error, another a vulnerability. Circumstances matter. 2030 ignores circumstances.
@algernon I agree that you can not compare a software error with an attack and the xkcd comic does not seem to me to make this comparison. This comparison is introduced in the article you mentioned. From my point of view, I was referring only to errors. But even the error of a single vote can make the difference between winning or losing an election.
@pepsy Indeed, but the chances for that are low, and if the voting machine prints the vote as a secondary paper trail, such errors are easy to catch and remedy. Counting votes is also a LOT simpler than whatever an elevator (let alone airplane) software needs to do. I'm pretty sure we can implement a counter bug-free by now.