lol ya, but you should look over the source.  it is interesting and may help

I highly doubt the mighty Notch himself would descend to speak with such lowly mortals as us, when there are so many other people who can probably solve our problems

eh? he spoke to SirCmpwn

"Hey Notch do you know Axe?"

no I mean what the code means

Again, you should be able to read most of it with very rudimentary C knowledge.
The problem is the same with any other random source code.  We don't already know the intentions of the code.  And it is large.
But i think it is quite doable.

Well one thing that I noticed is the createAndValidateTopMap() routine in the LevelGen class:
public static byte[][] createAndValidateTopMap(int w, int h) {
int attempt = 0;
do {
byte[][] result = createTopMap(w, h);

int[] count = new int[256];

for (int i = 0; i < w * h; i++) {
count[result[0][i] & 0xff]++;
}
if (count[Tile.rock.id & 0xff] < 100) continue;
if (count[Tile.sand.id & 0xff] < 100) continue;
if (count[Tile.grass.id & 0xff] < 100) continue;
if (count[Tile.tree.id & 0xff] < 100) continue;
if (count[Tile.stairsDown.id & 0xff] < 2) continue;

return result;

} while (true);
}
The map must contain at least 100 rock, sand, grass, and tree tiles, and at least 2 stairs, otherwise it generates an entirely new map and tests again. Note that result contains two arrays. result[0] is an array of map tiles, and result[1] is an array of level data (objects and whatnot, I guess).

Looking at this, I don't know how slow it might be to generate such a map on the calc...I guess it depends on the rest of the level generation algorithm to see how likely a valid map is to occur.

No, that would take away any sort of replay value. It removes the interesting aspect of a sandbox game like that.

Yeah, at least we could get all the stuff in, and worry about the map generation later

I'm more concerned about how it generates areas and pieces them together (so you don't have scattered desert in the middle of a forest, etc.)