Ink, Trust and the Electronic Vote

It probably passed unnoticed by most, but an editorial in yesterday’s New York Times contained this phrase:
Electronic voting machines that do not produce a paper record of every vote cast cannot be trusted.
The Times stated its support for Representative Rush Holt’s Bill which would ban paperless electronic voting in all federal elections. Of course, it’s the combination of ink and paper that supplies the level of documentation for which the congressman is looking. It is asserted that a physical manifestation of the vote is required to establish trust. A mark upon a ballot that can be plainly seen by anyone in the broad daylight of a town square.
While the documentation of voter suppression can be digitally captured and distributed via the real-time news network, the act of voting itself, apparently, cannot be trusted to the digital. The low cost of change damages the digital’s credibility here. It seems too easy to hack the vote. And yet, we trust our finances to purely digital systems— and our medical records will soon move from ink and paper to databases.
What would electronic voting have to be in order for it to enjoy the level of trust accorded to voting through the medium of ink and paper? And what change would that level of trust signal?