I think one particular line got me thinking:

  Graffiti happens at the intersection of ambition and incompetence: people want 
  to make their mark on the world, but have no other way to do it than literally 
  making a mark on the world.

Paul Graham has to make a caveat because it's too easy for someone to say, "WELL, sir, you have clearly missed that some graffiti is amazing. Banksy!" But there's a very important filter that should kick in: "Can I trivially amend his point in such a way that his argument still holds?" It turns out that I can, and PG should never have need to caveat his statement.

Let me propose, for what it's worth, that a "Banksy!" reply to a statement like the one above is the sort of accidental troll that invites the slow degradation into real trolls. Whether or not Banksy is graffiti or whether or not graffiti is art: these aren't the question of the submission. This isn't to say that tangents are always bad. It's rather that these sorts of discussions quickly turn in to "what color should we paint the dog house?"* That is: lots of firmly-held beliefs with little that can dissuade someone.

So: when posting why someone is wrong, first see if there's a small or trivial way in which you can 'fix' their point. If you can, their point wasn't really broken in the first place.

*I've long been trying to find the original 'why meetings go bad when you're talking about something everyone has an opinion on', but I can't. If anyone can help me out: much appreciated.