I'm not sure exactly what I want to say here, so I'll just dive in.
D&D experience points are primarily a challenge-based system. Players are rewarded based on the challenges they overcome, regardless of what that challenge actually represented in the campaign. For example, if they are trekking across the wilderness trying to find an abandoned ruin, they might get lucky and find it within a day; or they might get unlucky and take a week to find it. Assuming that the DM uses random encounters (or like me, ‘planned random encounters’, which I might talk about another day) to occupy the PC’s during the adventure, these two extremes represent very different levels of XP gain. At the limit, the group may be a completely different level when they hit the ruin than the DM expected, forcing him to adjust the encounter to present the ‘correct’ level of difficulty, where ‘correct’ depends on how challenging (or critical) the discovery of that ruin is to the campaign. This is despite the fact that the same goal has been reached in both examples: Find the ruins.
Over the years, this has become ever-more increasingly broken to me, because it makes adventure planning very difficult and can exponentially increase the DM’s workload, especially with published adventures which have ingrained assumptions about party level which are sometimes very difficult to adjust. XP is the primary reward mechanism for D&D characters, and defeating monsters is the primary way of awarding XP; therefore the question, “What constitutes ‘defeating’ a foe?” is perennially important. I always need to know when -- and how much -- XP to award the party every single time a foe is killed… or escapes… or is circumvented entirely. Often I have to make the exact same judgement calls discussed in the comments to the Didn’t We Beat You Already? post below.
More and more I find myself eliminating these problems by switching the reward mechanism to a goal- rather than a challenge- based system. In the current challenge-based system, the journey is more important than the destination. Traps, for example, have a CR associated with them, and it doesn’t matter whether the Rogue bypasses the problem with a single die-roll, or whether the unfortunate party whose Rogue is unconscious meticulously solves the conundrum with cute and lateral thinking… the XP award is the same. You'll even get parties who bypass the trap altogether, walking blithely around it via the secret door they uncovered five minutes earlier. In that case, the trap might never as well have existed for all the good it did in challenging them, or for all the role it played in their advancement.
Similar reasoning applies to NPC’s or enemies; when is their Challenge Rating actually relevant? Only when the PC's face them in a direct confrontation and when the combat abilities of the villain can come into play. No-one ever said a villain's CR should be increased because he has a good brain on his shoulders, or because he's adept at sending his minions to do his dirty-work... if for no other reason than that such esoteric measures are difficult or impossible to quantify in a cogent fashion. Imagine a physically weak but villainous mastermind with a network of implacable servitors, infiltrating the upper echelons of society and gradually gaining control of the multiverse. This is a worthy enemy for any group of adventurers, but as an actual foe he is worthless unless he has a set of in-combat abilities which can challenge the PC's at the climax of the adventure. Unless the DM makes him a potent wizard or psion (or whatever), then actually facing him down is irrelevent... it's the defeat of his plans that's the important thing.
This is where goal-based thinking improves on challenge-based thinking. Divide the adventure into story- or character-goals, even give them Challenge Ratings if you want. The fictional campaign against the asthmatic mastermind I mentioned above could easily be divided into a series of sequential goals: discover who paid for the hit on the mayor; bring the perpetrator to justice; uncover the double-agent in the Grand Council; break the villain's hold on the Pit Fiend; engage the demon's help to assail the villain's stronghold; destroy the mcguffin before the world's children become his slaves! In theory, every one of the goals could be accomplished without the need for any Challenge Ratings at all (even though in practice many parties get bored if they're stuck out of combat for too long). Inequities which may come into play between different groups' solutions to these problems become essentially irrelevant (to the DM at least), because they're just different ways to the same goal. XP awards are no longer invested in foes which the PC's may defeat -- or, critically, foes which they may bypass altogether -- but in the adventure itself.
A simple expression of this idea is to decide that this week's module will reward 1000 XP to each of the PC's. Each goal may be an equal part of that total award, or the goals may be weighted according to their difficulty or importance to the adventure. The how's and wherefore's of the party's approach to those goals is irrelevant, and as long as they have effective visibility of their current and future goals, they should never 'lose out' on XP awards simply by missing the secret door in the first floor privvy or stone-shaping through the wrong partition wall.
Like any XP system, this idea has its ups and downs. On the downside, a party which grinds a path the hard way all the way through the mastermind's minions might feel justifiably short-changed that their sweat and blood hasn't yielded any more reward than the party with the bard who charmed his maid into letting them in the back door (I exagerrate, but you get my point). Secondary to the goal vs. challenge issue, there also isn't a whole lot of room for circumstantial or individual excellence in this system, because that adds back in exactly the kind of unpredictability I want to avoid; so, the low-Charisma character whose player executes a masterful piece of in-character persuasion after the bard gets drunk probably won't get rewarded for pulling the party's ass out of the fire. On the other hand, apart from the dual advantages of predictability and not worrying about whether a villain has been defeated or not, a goal-based system encourages lateral and imaginative thinking if the party knows that the path of least-resistance won't get them any less XP, and encourages the use of all of a character's abilities at the table.
There's nothing new here, I'm sure there are many RPG's that use this method of player reward, but D&D isn't currently one of them (although some of the whispers about 4th Edition suggest things may be changing). One of the richest and most unique qualities of an RPG is its open-endedness, in that the players are free to think up any number of ways of getting from A to B. The XP system should encourage this kind of thinking, not strait-jacket the players into a series of combat encounters (no matter how fun) just because the most important reward in the game is invested in the means, not the end.