Ever since I first learned of its existence, probably in the pages of Dragon, I've been both fascinated and impressed by the Blackmoor campaign. This was, after all, the campaign where dungeons were born. Consider that for a moment. I often call Gary Gygax the Dungeon Master as a token of respect, but the truth of the matter is that, if it hadn't been for Dave Arneson, the game that evolved from the fantasy supplement to Chainmail would have been one without dungeons, which are Dave's greatest (but by no means only) contribution to the development of D&D.
What I really like about Blackmoor, though, is its scope. Unlike many campaign worlds, Blackmoor is small. The area covered by the map included in the First Fantasy Campaign covers an area of approximately 400 by 600 miles. The world beyond that area has never been described and that's because, unless I am gravely mistaken, there was never any need to describe it. Dave's campaigning over many years all took place within this comparatively small space, with a great deal of the action being within an even smaller subset of this area.
I frankly find that inspiring and, as I am learning, indicative of the best way to run an old school campaign: zoom in on a manageable area and describe the hell out of it. That means establishing the existence of lots of little settlements, not just the big "important ones," because, in a campaign like this, every settlement is important. The same goes for NPCs. There can be no nameless NPCs, not even the guy at the general store who sells you iron spikes and bullseye lanterns. He needs a name, a personality, and at least a hint of a life outside of his interactions with the PCs. All of that is the stuff from which future adventures can be written and are every bit as important as stocking your megadungeon.
The best campaigns I ever ran or played in had fairly limited scopes geographically. They all involved the PCs running around and doing things within a somewhat prescribed area, which was built up over time so that, by the end of it, we all had a very good sense of what the place in question was like: what it looked like, who lived there, and what went on when the PCs weren't making a mess of things. The trick, of course, is to be flexible and to let these details evolve through play, at least in part. Too much pre-planning and you run the risk, in my experience, of indulging in world building for its own sake. On the other hand, too little pre-planning and you might lose any chance of establishing a "reality" with which the players can meaningfully interact.
It's a tough balance to strike and I'm still struggling with it even now.