Polls, Redux

Looks like I'm not the only one thinking about the role of polls in elections today. This from Tom Tomorrow:

And yet. What does it mean when poll results vary wildly--when Bush has a double-digit lead in one poll, and yet the race is apparently neck-and-neck in another? Perhaps these polls have a twelve point margin of error, which is to say, they're complete guesswork. Perhaps one of these polls is trustworthy and the other uses questionable methodology--but the very fact that questionable methodology can exist takes us back to my initial point: I don't trust polls.