[throw5away]

I'd go as far as to say Approval > FPTP > IRV. FPTP has good properties like monotonicity (for a good example see http://zesty.ca/voting/sim/), easily distributed counting, and preservation of the secret ballot that IRV does _not_ have.

You really don't get much bang for your buck by switching to IRV; I'd say it's a negative deal overall.

[dane-pgp]

If we're comparing Approval and FPTP in terms of easily distributed counting, it's worth noting that a ballot paper with N candidates can be marked in N different ways with FPTP and 2^N ways with Approval.

That may not be a problem for computers, but if you want to avoid the potential failure modes of (and conspiracy theories related to) electronic voting machines, there is a big difference between an election official trying to place a ballot on one of 10 piles versus trying to find the correct one out of 1024 piles.

A system like MMP can avoid this problem by using FPTP counting of ballots and then topping up the number of seats for each party to make the distribution of winners match the distribution of total votes.

[throw5away]

You don't need the n bits of information from each ballot. You can just transmit the aggregate counts. A human can easily put tally marks in each of n buckets. In fact, a computer would likely do the same instead of creating an exponential number of buckets...

If you insist on a degenerate formulation, just split the ballots into n different elections and transmit the results of each one of those elections. We can use the same equipment we have today.

[alisonkisk]

Even simpler, an approval voter can cast up to N (different) ballots.

[dane-pgp]

I like this, although with paper ballots it would mean counting (up to) N times as many pieces of paper, which would be (up to) N times slower/more expensive.

[juped]

you cannot compare single-winner and multiple-winner systems at all