[at_a_remove]

Somewhat adjacent, I have been interested in generating mazes of the "cave" form. Many algorithms exist here, even some cellular automata approach for a more realistic simulationist approach. However, I have been kicking around ideas to take the very irregular results from that sort of tack and then abstract them back to a nodes-and-edges parallel data structure. Which is to ask, in this grid of filled in or empty cells, which gaps are the caves, and which are the twisty little passages?

At least for games, I am starting to lean toward the idea that humans like the simulationist approaches as "feeling right," but the abstract graph is better for the machines, and that the ability to start with one and then produce the other is where we will eventually go.