Set up an R&D program on internet voting with every expectation that it'll be implemented once it's ready to go. It's probably provably more accurate than other forms of voting and it would reach many more voters.
Internet voting's problem is auditability because there isn't any 'paper ballot' to rely on for re-counts.What do you think?
To see how important this factor is, take a look at what likely happened in Ohio in 2004, where there were no paper ballots, only electronic records. Obviously, nothing has been proven, but the anecdotal evidence - coupled with the high level of motivation on the part of Republicans - is strongly suggestive of electronic manipulation of the voting totals in many counties in Ohio.
Here in Los Angeles County, we use a numbered paper ballot marked with ink. As a former election inspector, I am aware that there are ways to possibly tamper with the vote totals, but since the counting at the precinct is done publicly, it would take a rather large conspiracy (amongst most likely total strangers) to effect a very minor (and readily challengable) swing in a precinct's vote total.
And those paper ballots, once their number is reconciled with the number of voters, are auditable and re-countable ad infinitum. The margin of error is really very small, as is the opportunity for manipulation.
I think that any internet-based voting system would still require a 'machine' in a central location that would print out a paper ballot. Otherwise, your vote exists only as a small pattern of bits stored in a database somewhere.
And as a database developer and data analyst, I can state authoritatively that you do NOT want people like me (no, not me specifically, but people with my capabilities, less my ethics) anywhere NEAR your pattern of bits in a database.
Having said all that, it could be done, and would be VERY cost-effective: a system like this would scale up quite nicely, and with voting occurring over a long period of time - say a month - the system wouldn't have to be awfully powerful or consume much bandwidth, either.
Of course, this is WAY to logical a solution to our rather creaky - and error-prone - system.
Suffice it to say that any election decided by less than 2% is well within the margin of error, and beyond any possibility of conclusively determining a winner.
Vote on public policy ideas and identify like-minded candidates in 2012 elections at the99vote.com