On Monday 27 September 2004 21:33, Enrique A. Chaparro wrote: > On Fri, 24 Sep 2004 10:01:59 -0400 > > Second objection is: > Even if you were able to overcome the first objection (wich, in case > "b" above, will solve the issue for simpler methods), a secure system > requires _huge_ amounts of computing power (if you're thinking of large > scale elections, i.e. elections involving millions of voters) > > Regards, > > Enrique There is no need, and I believe it undesirable , to have every individual vote tallied at a central site (a mirror , disaster-recovery, site is not solving the problem in question). A hierarchical system is appropriate. furthermore while exit polls influence the election and so on we also do not want local partial results disclosed. It may be desirable for updates on a periodic basis to be forwarded as a 'sign of life'. The problem is that if these intermediate results exist news agencies will want to publicize them and that would influence the election in-progress. Already we have seen partial dsenfranchisement and lower voter turnout in Alaska and esp. Hawaii since the polls are open there after the polls in Eastern US close. On a technical level each voting machine should tally its results. The next level up is the Election District. Results of each voting machine are summarized by the ED node and reported up to the location master node for the polling place (again I leave out redundancy; we assume measures are taken for reliability). The results of each polling place are grouped by county. It is not clear to me that any state-level elected office has a district spanning multiple counties. Similarly I do not believe that any congressional seat spans state borders. Therefore it should suffice to tabulate state-wide and federal elections on a county-by-county basis for each state at a state-wide master node. In short, as has been pointed out a tree approach makes the problem of vote counting quite tractable.
