Well, Mark and John and Ed responded to my three points and answered my question and more. So, in no particular order, here's a few thoughts on the issues they raise [looking back over it after writing, I think the debate probably boils down to #'s 4, 6 and 7]:
1. Let me make one big concession at the start. I wouldn't vote for a law than banned polls, and were I dictator of the universe I wouldn't want to outlaw them. (I'm a fan of free speech; I'd get that first amendment tattoo in a heartbeat; etc.) The position that I would feel more comfortable defending is something like "polls are on balance a bad thing." Or, even more milquetoasty: "polls are not the best use of newspaper resources."
2. I hope that goes along way towards addressing the main thrust of Ed Kilgore's argument -- namely, that I am weirdly in favor of reducing the total amount of information that exists in the public sphere. I think it's fair to say, with Ed, that there should be a strong presumption in favor of putting additional information in the public sphere. But I think that's very different from saying the current level of information is optimal or desirable.
3. I think there is an easy extension of points #1 and #2: Producing information isn't free. A dollar that the New York Times spends on a poll is a dollar that it isn't spending on a Baghdad bureau or a congressional beat reporter. Well, that's not really how tradeoffs work. But the general point -- you have to measure something against its opportunity cost -- is one I agree with and one that is relevant here. So one question to consider might be this: Could the resources that the Times and the Post spend on polls be better spent elsewhere?
4. I am happy concede John Sides' empirical point that there's little evidence that politicians are "buffeted to and fro by the winds of the capricious public," as expressed in polls. I am also willing to concede that politicians do in fact use polls "to find out how best to sell their ideas." But I'm not sure what this proves. If politicians become more adept at selling their ideas to the public, does that benefit the public? That sounds like a question with a complicated answer. On the other hand, I feel reasonably confident saying that it is not the purpose of the media, at least as traditionally understood, to provide a consulting service by which politicians can hone their rhetorical skills.
5. John Sides also has a lot of helpful stuff to say about information cascades, but I'm not sure what to make of his conclusion -- namely, that "the role of polls [in influencing future polls and outcomes] is unclear." I agree that it would be difficult to prove how a poll in the present affects a poll in the future. (Or at least I'm not sure how you'd ever tease causation away from correlation.) But it would surprise me if they did not. We all have preferences for all sorts of things that are popular -- neighborhoods, fashions, music and so on -- which will in turn make them more popular. These preferences are endogenous or interdependent or whatever you want to call them. I think it would be strange to discover that political preferences are any different.