"The milieu for initial adventuring should be kept to a size commensurate with the needs of campaign participants -- your available time as compared with the demands of the players. This will typically result in your giving them a brief background, placing them in a settlement, and stating that they should prepare themselves to find and explore the dungeon/ruin they know is nearby. As background you inform them that they are from some nearby place where they were apprentices learning their respective professions, that they met by chance in an inn or tavern and resolved to journey together to seek their fortunes in the dangerous environment, and that, beyond the knowledge common to the area (speech, alignments, races, and the like), they know nothing of the world. Placing these new participants in a small setting means that you need only do minimal work describing the place and its inhabitants." (DMG, pp. 86-87)I daresay this is the source of the whole "you all meet in a tavern" cliche, even if its original origins may have been mislaid over the years. The section continues to describe the method by which the campaign world is slowly detailed; the basic gist of it is that the world only needs to be described as far as the player characters need it to be. The dungeon master need not flesh out the world until the players are likely to need it fleshed out. This is the bottom-up approach to setting design. You start with a village and a dungeon, and you work from there. It's an approach that should be well familiar to those of us who cut our teeth on the Village of Hommlet, long before there was any Temple of Elemental Evil to move on to afterward.
Back in 1979, this approach made sense. There were no published D&D campaign worlds other than Judges Guild's City State of the Invincible Overlord, and that fit into the "detail a single area, and move out from there" approach. There was no grand overview. No sense of how the City State of the Invincible Overlord fit in with the City State of the World Emperor, and how either of those political entities interacted with the Barbarian Altanis or what the heck were the Isles of the Blest. Isolated spots, like the city-states, were incredibly well-detailed, but the grand overview was lacking. It was a world built of pinpoints, built from the bottom up. (The way the wildernesses were described was singularly appropriate for what we now call sandbox play, but that's beyond the scope of this post.)
Of course, with the publication of the World of Greyhawk Fantasy Setting, that changed. Now there was a coherent framework, built from the top-down. You knew where the City of Greyhawk was relative to Verbobonc and Urnst, and had some idea of the politics and factions and races, and how they all fit together. When the boxed set came out a couple years later, the picture was even more complete, as it saw fit to include gods and goddesses and much more detail.
I would say that the initial publication of the World of Greyhawk Fantasy Setting folio represents a watershed in campaign design theory. It overturned the established order. No longer was a village and a dungeon enough to start a campaign; now you needed at least a continent, or a couple of kingdoms at the most bare-bones level. Gary Gygax himself established the new theory of campaign design in 1983, in an article in Polyhedron magazine entitled "Setting the Milieu".
"Rather than working haphazardly from a dungeon to a world in ever-widening concentric circles (and I have done this in the early days), the DM should broadly outline his or her universe, sketch out the world upon which initial action will occur, generally detail a continent, develop a section of that continent (perhaps four or five states), prepare a full history of the central area in which the adventurers will find themselves, and then begin recording the minutiae of the campaign. Highly detailed work must be done for the major urban and rural settings of the heart of the action." (Polyhedron #12, p.16)What a difference four years makes!
I must say, once the World of Greyhawk folio came out, I don't think I ever even attempted to run a game that didn't have a continent-wide map (or at least a sub-continent). Until a couple of years ago, that is. When I decided I was going to run an AD&D game, I put together a town, a dungeon, and a quick wilderness map. That was it. After a couple of gaming sessions, I plopped the whole thing into the World of Greyhawk, but only out of love of that particular setting; I could very well have stuck with the original conception and see where it took me.
That, I think, is the heart of the difference between the two approaches. The top-down vision of campaign design gives the dungeon master the final say in the broad brush-strokes. The kingdoms, the religions, the planes of existence, are already at least sketched out. Sure, they can be changed, but even the act of putting them down on paper gives them a certain permanence. The bottom-up approach, on the other hand, might be seen as giving the players a certain hand in the development of the setting. Your dwarf fighter wants to be a disinherited prince from a dwarven kingdom? Well, you could try to shoehorn that into an already-established setting, but if the setting consists of a town and a dungeon, then your options as a DM are a bit more broad. This speaks to the "say yes" philosophy of DMing.
My current game is on hiatus for a variety of reasons. I wonder if it wouldn't be nifty to restart it with a bottom-up approach. What do you all think? I map out a village, a 30-mile wilderness, and a dungeon. "You all meet in a tavern, and decide to go to the village and explore the dungeon."
Then just say "GO!"