I support IRV as a substantial improvement over plurality-takes-all; but IRV is not "perfect" or even the best we could hope for. Here are a few reasons why:

1) IRV does NOT guarantee a majority winner. I don't know why people continually say it does-- it clearly doesn't. Imagine a three-way race like the following, after one round of voting:
Liberal: 10 votes
Moderate Conservative: 9 votes
Radical Conservative: 5 votes
Under IRV, the Radical Conservative is eliminated. Now suppose that 2 of the Radicals chose the Moderate-Conservative as their second choice, but 3 didn't specify a second-choice. That would give the Moderate 11 votes, beating the Liberal, but NOT a majority (11/24 isn't more than 50%). This is NOT at all implausible, and in fact, would be quite likely in a 9-way primary race, particularly with IRV proposals that only allow voters to specify 3 choices.

2) Some proposals of IRV do not allow full expression of voter preference, and thus, encourage "lesser of evil" voting and "spoilers." In Michigan, there is a movement for an "IRV; 1,2,3" voting system under which voters could specify their top 3 choices. I think back to the Democratic primary-- my top three choices were Kucinich, Moseley-Braun, and Sharpton. If I voted for those 3, there is a very good chance that my vote would in fact be "wasted." To stop this from happening, I would have compromised my vote to 1) Kucinich, 2) Mosely-Braun, 3) Dean. Compromising votes is what's wrong with our current system, we shouldn't repeat it in IRV.

3) It doesn't really address the spoiler problem as well as it could. Imagine I really liked Nader, but most Kerry supporters preferred Bush to Nader. Imagine enough people like me ranked Nader ahead of Kerry to make it a 2-way race between Nader-Bush. Now imagine that enough Kerry backers supported Bush over Nader to give Bush the win. If defeating Bush were really my top priority, and I had the foresight to see this might happen, then I just might rank Kerry ahead of Nader afterall, in order to prevent a Bush victory-- which is exactly what people are doing under the current system.

These are 3 serious flaws with IRV. A superior form of ranked-choice voting is known as Condorcet Voting, but the vote counting process is an algorhythm that would require computers, and we all know how dangerous electronic voting can be.

There is a simpler solution than Condorcet that is much better than IRV. It is so simple, in fact, that many people dismiss it without giving it proper consideration. It is called "Approval Voting," by which people can simply vote for as many candidates for a given office as they approve of. There are no promises of a majority winner, but these promises under IRV are false, anyway. What AV does guarantee, though, is that people will never ever be afraid to vote for the candidates they actually prefer, while they may indeed choose to vote for the "lesser of two evils" as well. Instituting AV would be much easier, and would show an instant appreciation in third-party support. Third parties doing well in the polls would cause people to do more research on varying viewpoints, thus saving the world from inevitable destruction.