This is a very thoughtful counterpoint to my question, and deserves more testing. I do wonder how this compares to voters filling in the wrong bubbles on optically scanned ballots, and other errors introduced to a process that produces something less readable than a BMD.
I remain open to persuasion on the issue, based upon what evidence we see in the field. I'm quite happy to see DRE machines replaced, even with BMDs---almost anything with a paper trail is a great improvement.
And I have no dog in this fight, other than professional interest in what is the best system balancing all kinds of things, including security, integrity, and access for voters with disabilities.
