I do agree with Raph Koster's recent statement that narrative is not a game mechanic. But that is not why you should read that post. The reason why you should read that post is that Raph presents this brilliant model of games, in which every game is a sequence of problems, black boxes in which the input of the player is processed, and feedback.
That model explains so much better some of things I don't like in modern games, for example how raiding is just like Raph's "racing game" example: The cognitive problem of what to do is small, and the interface problem of doing is correctly is large. Or some things I do like, but recognize as a problem, namely games that are full of story and thus have lousy replayability.
Thus I consider Raph's post a must read for anybody discussing games. Check it out!