Espeically if, like in minecraft, you can set your own seed - perhaps there's an interesting social graph problem in automatically assigning the same seed to clusters of friends but keeping them somewhat random globally.

I thought about that but I remember when one of the great things about this sort of gameplay was having a friend over and being the one in-the-know for all of these little secrets and helping each other out.

But then it loses something, and you end up with Diablo. Not that Diablo is a bad game, but it's a decidedly different kind of game than Zelda.

Right, that's what makes it a classic hard design problem. The best solution, to the extent that it even exists, is necessarily going to be imperfect, and will be found somewhere in the middle.

There is no "best solution" or even a "solution". IMO games are art, so trying to solve for a solution is a fool's errand. Even games like Call of Duty -- they're pop art, which is simply the artistic process as democratic exercise. Zelda, for many of us, represents that child-like wonder that we can never return to for the same reasons game developers could never make a similar game.

I don't think we're disagreeing. There's no perfect solution; it's like designing a car or directing a film.

Though I suppose I disagree with your assertion that one can never recreate or improve upon anything that evoked childlike wonder. Portal 2 did that for Portal; Mega Man 2 and 3 did that for their prequels, Empire did it for Star Wars, and as someone with powerful childhood memories of the mid-1980s Main Street Electrical Parade, I've been hearing good things about Disney's new Paint the Night.

sure when appropriate randomize it in clever ways, but don't make it random unless it really should be random, cleverly placed secrets are part of why many players like these games, internet spoilers or not. See Myst.