The personal computer (PC) has more than enough capability to guide a voter through a screen-based ballot, to prevent overcounts or undercounts, to conform to all voting rules, and to allow voter changes until he/she "pulls the lever" and receives a filled-in paper ballot.  Two million PCs, each handling about 50 voters is about the number required for a national election.  The 600 to 700 minutes or so of "poll's open" allows over ten minutes per voter and 20% spares.  Here is a ballpark cost estimate of the dream system.

Two million PCs with capabilities far in excess of what is needed can be purchased today, at retail, for less than $1.6 billion. An agency buying in such quantity should pay less than half.  Very sophisticated system software costs no more than a few tens of millions of dollars.  Special software is needed but would not add significantly to the ballpark estimate.

One printer for every four or so PCs adds less than 10% to the cost.  Ninety-nine percent of the PCs would be in the precinct voting places. Beyond standard election costs – space rental, supplies, maintenance, staff and official pay -- there are no further costs.  The total cost of the proposed voting system should be around a billion dollars, much less than many would expect.

At the close of any election, all votes need to be counted.  An average of ten PCs  and two or three printers in a precinct need to be networked to produce a precinct tally record.  The precinct vote counts pass up the voting chain to about 1% of the PCs distributed mostly at the municipal level, some at the county level, and 50 at the state level.  The links of the voting chain are the precinct, municipal, county, state, and national levels.  Some large cities have sub-municipal levels.  House districts cut across county lines and are totaled at the state level.  Media will obtain outcomes at all levels from municipal on up.

National media will want tallies as soon as the polls are closed at the municipal level if that is earlier than the state poll closing and at the state level as soon as all polls in the state close.  In the three evening hours or so, as polls close from east to west, state by state, the media will report on the congressional and gubernatorial results and announce the running tallies both electoral and popular votes for president and vice-president.

Exit polls, which contributed to the 2000 fiasco, will be irrelevant. Counting votes is simple addition for each candidate aggregated at most five times as the vote counts go up the voting chain.  This can be done accurately and reproducibly with today's technology in a few minutes, each uplink taking as little as a fraction of a second.  Each PC has only a few kilobytes of data that needs to be passed up the voting chain.  After the last precinct in the nation closes, rapid tallying can be done for the final count or for a recount with complete accuracy.

These processes are similar to the way that a large financial company with hundreds of branches can determine and confirm overnight cash balances accurate to the penny.  Communication channels are required for at most only a few minutes at the close of the election and they need carry only a tiny bandwidth.  There is no need to use the Internet, avoiding exposure to bugs, denial of service, encryption/ decryption and viruses.  You can bet that thousands of hackers would lick their chops for a shot at what would be a biennial jackpot, corrupting a US national election through the Internet.

Yes there are special processes required for properly handling absentee ballots, write-in votes, and dubious registration categories (ex-felons, address changes, etc.).  These are discussed in the sequel.  One basic idea is that no voter need be turned away from a voting place because s/he appears not to be registered.  Many could vote in a category whose tallies are separately and coherently handled as they are carried up the voting chain.  Category vote legitimacy need not be resolved if the counts are too small to affect the outcome.  Two major components of the best-practices system which help handle registration problems are the resolution of the "too-close-to-call" situation at the state level and a special process for remedying what is called the "sanctity of the vote", that makes fraud by any voting officials from the precinct level up to and including top state officials, very unlikely and rather easily spotted.

Updated 6/14/2008

It never occurred to me that the US government would shell out (or require states to shell out) a major $2 billion investment in election technology, based on pc’s and new computer software, without owning the software and source code required to prepare and confirm the variety of ballots required by our annual nationwide voting system.

What has happened in the last seven years is that any company could offer touch-screen, direct recording electronic, or other voting systems, whose software, including source code and hardware, were proprietary and jealously concealed by the company, and whose inadequate technical and maintenance people might or might not be available to handle breakdowns or loss of correct vote-counts with equipment too tolerant of error and too dumb to resolve ambiguity. Any company that met minimal standards has been free to sell its voting systems to states that often over the years have accepted and had to operate the upcoming election with sub-standard equipment based on company promises for improvement often not met in time to conduct the mandated fixed-date election with alternatively available equipment.

The cost of dealing with these errors has been many times the cost of the “Best Practices System” #13.1 estimated at under $2 billion in the first year and far less thereafter. This stupidity has cost our country billions of dollars.