To set things up, let's get a definition in place. For the purposes of this discussion, metagaming is the use of player knowledge to make decisions for their character that the character wouldn't otherwise make (because the player knows things the character doesn't).
Traditional wisdom says "metagaming is bad, always", and a lot of traditional style RPGs tend to assume this - character knowledge and player knowledge is separate, and the player shouldn't use his own knowledge to get an advantage.
Thing is, original D&D didn't take such a stance - from my research into the subject. It was a wargame (as RPGs didn't really exist at that point as a discrete medium), and the character was a playing piece, little different to the models in a wargame army. Player knowledge - and player skill - was important, because knowing what to expect had a huge impact on success, and learning the risks the hard way (through dead characters) was an assumed part of the game.
On a more modern side of things, there's an increasing trend of games that handle players more as authors than actors. An actor-stance game is the traditional style, where player and character are synonymous, and the player doesn't make any decision that couldn't be viewed as a character decision. An author-stance game has the player in control of the character, but the player is 'external' to the character, making decisions as if writing about the character. Author-stance games often pass a degree of the world-creation and narrative editing power from the GM to the players, and often give the players some benefit for making decisions that aren't in the characters' best interests.
Over the last few years, my tastes in RPGs have shifted more towards the author-stance, where the game acknowledges a difference between the player and the character, and metagaming isn't such a taboo because they're built around the idea that players will make decisions for out-of-character reasons.