I started work on a second version with a significantly nicer code-base and a bitmap based level editor. Meaning user-created levels would be fairly trivial. The hard part of that was automating the detection of adjacent pieces. I had an algorithm for doing it but there were a lot of “edge-cases” (hah, pun) that seemed like they needed human decision-making. I ended up just leaving the project in an unfinished state.
However, these kinds of threads popping up really make me reconsider making/releasing a “squeal” of sorts.

An interesting variant problem is where for each region you’re given a list of four colors you can use for that region, with the same restriction that the color you pick for each region (from its respective list) can’t match the color of any of its neighbors. It turns out that putting four colors on each list is not enough–that is, there are some maps and list assignments that have no solutions. But with five you can always do it.

The problem of finding such a map where this is impossible was posed several decades before it was solved by a seventeen-year-old. So maybe Terry’s secondary school efforts weren’t entirely in vain.

That’s an interesting point about the torus. I’m currently toying with the minimum number of colors required for a lobachevskian rather than euclidian map. I’m pretty sure it’ll be 4 since you can model the one as a subset of the latter, but what else am I really doing this evening?

Man, I spent a lot more time on this game then I thought I would. “Oh this game seems kinda simple I don’t see why I’d keep playing it hey what the hell more levels please!”

(At one point in secondary school I was convinced that I’d discovered an obvious proof to the four colour problem. I suspect I was overlooking something, though.)