Classic RPG Design and the Art of Exploration

Responding to GameBanshee's What an Old RPG Can Teach Today's Designers editorial, Rampant Coyote has a few thoughts of his own.
Exploration is a big deal - perhaps not a requirement of the genre but certainly a hallmark of the best games in its short history. But too often in modern RPGs it is extremely limited, or ... unsatisfying. What was the difference?

Worlds were somewhat easier to make back then, due to the predominantly tile-based approach to world construction. As a developer, the tiles - while boring and repetitive - really constrained the action so we didn't have to worry about physics bugs, getting stuck in the geometry, pathfinding problems (well, not as much), and so forth. The difficulty of getting all these things to work together in a freeform world is daunting - whether in 2D or 3D. 3D worlds are just exponentially more challenging.

A similar issue with exploration comes with the difficulty of creating content for modern RPGs. When a single dungeon takes so much time and effort to construct (as opposed to the tile-based days, where a designer could whip up a 10 x 10 grid with a handful of interesting encounters in a few hours), do you really want to spend so many man-months of effort and expense on optional content that only a handful of gamers will ever see? Theoretically, the old 2D RPG game designers could have built their worlds pixel by pixel. Come to think about it, I think Bioware actually did just that, with Baldur's Gate series. But it wasn't just a limitation of technology, I feel, that stopped the hand of most RPG designers from attempting this. It was also judicious application of available resources.

Both of these issues, in my opinion, could be resolved by adopting the idea of less granular building blocks for the 3D world. Yes, this means more repetitive pieces of content, arranged with tighter constraints, in a 3D world. However, the Elder Scrolls series, the Neverwinter Nights series, and I'm sure many other modern CRPGs - have had their own approaches to this very idea. But I think a lot more could be done here.