Earlier, Monte Cook wrote something about "crunch vs. fluff" being a false dichotomy. And other people weighed in, including Robin Laws. And I read these posts, and thought: "What the hell? Do you guys even know what the argument is really about?"
There are certainly minor disagreements between those who emphasize mechanical details and strategies versus those who emphasize fictional elements, either in characters or in background. But when people complain about fluff, in my experience, it's usually useless fluff, stuff that not only does not produce a mechanical advantage, but doesn't produce any interesting roleplaying. When people complain about the fluff in a rulebook, rarely are they complaining about stuff like "the city is secretly run by vampires"; they're complaining about having to read huge mounds of crap about which streets are paved with red brick and which are paved with gray flagstone, or the differences in weave patterns in the clothing of the various social classes. Or maybe they're complaining that the setting detail, instead of being in terms the reader can relate to, is instead written from an in-setting viewpoint, using special alien vocabulary, which includes a glossary in the appendix that readers have to refer to frequently just to decipher what's going on in the setting so that they can decide whether they like it or not.
You see, it's not the absence of mechanical impact that makes fluff a problem. We all know that there's such a thing as non-mechanical resolution, what Ron Edwards referred to as "Drama resolution". We know that if an RPG supplement mentions that the walls of a house are wood instead of brick, this can be significant, even if we don't have mechanics for how easy it is to destroy one compared to the other. The players could say "we burn the place down," and the GM knows that the wooden house will burn easier. But if the RPG supplement goes into detail on different brick-laying techniques, it doesn't really have an impact on play. The GM will probably not even use it; it's just a wall of text the GM has to read through to see if there's some relevant fact buried in all that crap. And worse, the GM will know that the writer is being paid by the word, or is otherwise expected to guarantee a product of a certain size, so all that extra, useless information was added as padding.
Or consider character details. "My character is a barbarian from a tribe of sea raiders" may be fluff, but that's not what people are complaining about when they complain about fluff. "My character likes butter," however, would probably be a good example. Because the "barbarian" factoid could impact play in many ways, creating fictional conflicts or allowing recognition of treasures taken from that culture; it creates events in play that the players can deal with. But the "butter" factoid, if it comes up in play at all, only creates butter-related events, which interfere with what the other players want to focus on. Sure, once in a great while, such a useless detail may turn into a useful detail, like a situation where the character needs to slide through a tight spot and comes up with the bright idea of using all that butter he's carrying as a lubricant. But the rest of the time, it's usually used as a bludgeon by one player to force other players to pay attention to him.
That is the question you should always ask when writing down a bit of fluff on a character sheet or in a gaming product: is this actually going to produce interesting situations in play? Or is it just going to draw attention to itself?