I think the division is between mission-style-play and character-motivated-play.
In the former it makes no difference, if the GM owns the NPCs, because he presents the players with tangible problems, that exist outside of their characters backstories (although said backstories might be what gets them into these "missions" in the first place). The latter, a style I have come to enjoy, is based on the assumption, that the players and the world are interconnected and what happens to the former, also happens to the latter: if a member of the thieves guild is found guilty of assassinating the king, than the character is removed from power as head of said thieves guild. The character concept changes from "head of thieves guild" to "ex"- a significant change. Now the player would be right to bitch about this, if he had no way to affect the outcome. He wouldn't bitch, if he wouldn't have been head of said guild- because the assassination of the king wouldn't have
affected his character otherwise. Heck, there is a reason of GM-advice cautioning against having high-level-thief steal all the equipment of the PCs while they sleep, because even in the former mission-style-approach
the equipment is considered part of the character (even though its really part of the world).
Now, all the above-mentioned examples are examples that
could happen in the world, but (usually)
don't for a
reason. And this reason is that (in most RPGs) characters are the players only tools for affecting the game world. But there is a difference between theoretical freedom of action and real freedom of action. If you
could take any action with your character, but wouldn't affect the outcome, than the GM isn't running a RPG (since RPGs are defined by the
interaction of the participants). The choice is illusionary.
So the players has limited resources (his character) to achieve certain results in the game-world (like becoming head of the thieves guild). The GM has unlimited resources (the whole world) to undo anything the player achieves (or simply grant it to him, which is just as unsatisfying). A good GM tries to engineer his resources in such a way, that they are balanced, allowing the player to tip them in either direction with his character.
To come back full-circle:
player goals and
character goals might differ.
What a character wants is not always what a player wants (or what a player wants
right away). So being aware of
player wants and adapting your campaign world to suit them.
"We want to slay the dragon plaguing the countryside." might mean anything from: "We want to get into a highly tactical dungeon, where we need our wits to survive and have thrilling fight at the end." to "I want my character struggling to overcome his fear of dragons and win the respect of his father." or even "I want the damn dragon gone, because I don't want to risk losing [insert NPC name here] to him. It is an annoying monster anyway." and the adventure would play out in a decidedly different way (i.e.: the second goal would be defeated, if the whole adventure takes place in an underground complex, where no one can see how the character behaves and where he can't interact with anyone, with the third goal it would simply annoy the players, if taking the dragon down takes way to long).
Being aware of these expectations goes a long way to adapting your campaign world, but frankly, its impossible to be aware of the all the time. People screw up. Losing is part of the roleplaying GAME. The key is being flexible enough to incorporate the ideas of your players and not sticking to the "I said it, its final"-shtick.
That's all.
Also see
here (mainly the posts by Daniel Wood, but the whole thread has some good observations).