"The fact that breaking windows would make a society poorer (fewer windows) is precisely why nobody ever proposes stimulating the economy by deliberately smashing windows. But the way the dialogue works is that first a Keynesian observes that fiscal stimulus can increase growth in a depressed economy. Second, as an attempted reductio, a conservative says “if that was true, then you could increase growth by breaking a bunch of windows.” Third, the Keynesian accurately points out that you could, in fact, increase growth by breaking windows. Fourth, the conservative accuses Keynesians of wanting to break windows or believing that window-breaking increases wealth. But nobody ever said that! The point is that we have very good reasons to think smashing windows would be a bad idea—there’s more to life than full employment—and that’s why Keynesians generally want to boost employment by having people do something useful like renovate schools or repair bridges."
...or colonize and terraform Mars.
Anyway - getting back to the point. The argument can sometimes get a little more complicated than this, because some people say that the act of government spending itself is the window-breaking. Then you have to ask why? The weak arguers will talk about taxing and spending at which point the discussion just needs to end. The somewhat smarter arguers will talk about crowding out in the bond market. At that point you say that yes, crowding out may very well be a problem at some point, but it's not now, so this window you think is being broken actually isn't being broken at all.
Yglesias also makes a reference to reductio ad absurdums. I personally think that reductios are often very bad economic logic. Economists think on the margin and they usually think there are lots of diminishing marginal returns to different things. Reductios are usually started in a way that completely ignores this. In truth that means they're not even actual reductios, because the reductio is supposed to carry an argument to its logical extreme, and if you ignore things like diminishing marginal returns or if you ignore the fact that we're making an argument on a particular margin you're abandoning the underlying logic of the claim, and therefore you're not really offering a refutation of the claim.
UPDATE: Bob Murphy points out that while Keynesians and Bastiat agree on the stock vs. flow question that anti-Keynesians regularly bungle in making accusations (well - he doesn't use those words exactly), Bastiat was pretty clear about assuming complete crowding out in his essay. In that sense, we can just say "Bastiat made the right point, but he messed up on the crowding out part... or at least was declarative enough that his parable can't be applied under all circumstances". A better rendition is given by Wilhelm Ropke, who doesn't make Bastiat's broader crowding out claims when talking about the impact of disasters. Brad DeLong has made the argument in the past that Bastiat understood there could be crowding in by public expenditures during depressions. I'm not entirely sure I agree with Brad on that, but at the very least he demonstrates that Bastiat supported a sort of "demand smoothing" by the state, which is still better than a Ricardian equivalence-esque non-effect.
UPDATE 2: The point is this - nobody flings Bastiat because they are having a disagreement about crowding out, which may be a point that Krugman and Bastiat disagree on. They fling Bastiat because they think that Krugman and others are somehow ignorant of the unseen costs of these things or ignorant of the concept of opportunity cost. If that's what you think, don't expect people to be impressed.