Would you consider a Linux port ?
The most frequently asked question about Cartographic Revision has been why it has such particular restrictions on player count. In order to answer that question, we need to discuss how Cartographic Revision works. My reason for making a new map generator has been because the default map generator's variance often overshadows players' decisions during the course of play. In order to control (but not eliminate!) such variance, Cartographic Revision takes a structurally different approach to map creation. This results in different limitations (and different opportunities) compared to the default map generator.
The main reason to have a map generator at all, rather than a static map, is to introduce variance: to present different challenges and opportunities from game to game, encouraging adaptation over rote memorization. There's nothing fundamentally wrong with such variation. Generated maps go a long way towards keeping Dominions fresh and entertaining year after year. The issue is excess. Some games can survive and thrive with enormous variance, but they're often short and/or single player games. Too much variance privileges random chance over player choices. That can be fine for a short laugh, or a for fast-paced reflex-based game, but it's a gigantic feel-bad problem in a deliberate multi-month multiplayer strategy game like Dominions!
Unfortunately, the default Dom4 map generator includes too much variance. When one start location has only 2 adjacent provinces and another has 7 adjacent farmland, the former is arbitrarily crippled while the latter arbitrarily boosted. When start locations share cap circles due to cramped placement or buggy cross-continental province connections, the weaker early game nation will almost certainly be smashed without recourse. When a player picks a nation with important terrain-based national recruits or summons and the only terrain for them is across the entire continent, their choice of strategy is randomly crippled. And when a player starts near all the thrones required to win, or has vast swathes of uncontested expansion room, the game can easily end without a majority of players having any of their decisions affect the outcome. I've experienced all of these, both when it benefited me and when it destroyed me, over the course of the past few years (which, because of the length of Dom4's games, is still less than two dozen games). Even when I benefited, the extreme swings of luck on turn 0 undercut my satisfaction with the game.
Screenshot from early province connection testing, illustrating the grid-based nature of the map. Red provinces are start locations.
Cartographic Revision addresses these problems by using nation choice as the algorithm's starting point. Looking at the stated order of operations in the default Dom4 map generator, it's fairly clear that it starts by creating a noise-based world and then tries to find appropriate places to draw provinces and place players. Cartographic Revision takes the opposite approach by placing players first and building the world around them. The map's core structure is a distorted grid of national start locations, which is half the reason for player count restrictions. The other half is because only the most squared-off options provide the feel of a normal dominions game, so I included only those size options (something I'm now reconsidering). Once nations are placed on the grid, and the grid distorted to discourage reliable rush play, the land radiates outward from each starting location. National terrain and rough terrain come first, near start locations, while high-value terrain like farmlands come later, between start locations. Provinces themselves compose a grid with frequent diagonal connections, though I've gone to some effort to visually disguise it. Throne placement is handled by a throne grid offset from the starting location grid - meaning each throne is spaced roughly between 3-4 players depending on the particular grid distortions. This controlled structure removes the most severe game-crushing variation, since no one will have either a crazy good or crazy bad start and will have at least minimal access to their nations' terrain-dependent tools (specifically: at least 1 province of the necessary type somewhere within 2 spaces of their start province).
Improvements aren't limited to fixing the default map generator's blatant problems; a gameplay-first map generator can also subtly encourage engaging patterns of play. In particular, placement of blocking terrain and high-value terrain can combine to create tense decision making. It's not interesting to have a totally impassable border, but it can be very interesting indeed to have a border divided into multiple fronts. The placement of blocking border terrain, like rivers and border mountains, generally flows outward from the starting locations (albeit with quite different algorithms between the two). This divides a player's expansion options into pie slices - sometimes very wide and sometimes quite narrow, but always with impediments to movement between slices. This means that army deployment can be a very serious commitment indeed. With high value provinces near natural political borders, players will be greatly rewarded for picking their fights wisely - and swiftly punished for misjudgment! Frequent fords and passes still permit movement between pie slices, often quite freely, but even a single impassable province border can make a profound strategic difference. Of course, this also rewards players who use their nations' terrain mobility to the maximum. If you can freely cross rivers or mountain passes, it's to your benefit to fight wars around those terrain features.
Hand-colored "pie slice" visualization.
While variance is reduced in Cartographic Revision, it is far from eliminated. There would be no real point to a map generator if it only spat out the same fundamental geography each time. While pie-slicing and start location balance are consistent patterns, the algorithms are untamed enough to produce a wide spectrum of specific results. Each of your political borders, and the paths between them, are likely to end up quite different in character. One region may contain unappealing deserts and mountains, while another might have terribly tempting farmland just a step over your neighbor's border. You might have a wide swath of open plains in one direction, while in the other lies a tangle of rivers and lakes. Playtesting has shown that choice of expansion path is enormously important on these maps because you can choose which pie-slices to embrace - and which to shun! - when building your early political borders. You can also manage your risk: play it safe by expanding in a controlled circle, accepting the lower average province quality, or aim for the sky by rushing straight to high-value provinces while leaving only a thin buffer around your capital. Early game power differentials still exist, but they are much less the result of the map generator and much more the result of how players choose to play their opening hands.
Most of the limitations in this new map generator stem from it's different structural approach compared to the default map generator, but I'd like to think it's a good tradeoff. Player count is definitely more limited. Map water percentage is tied purely to the number of water nations chosen rather than a separate setting. Provinces per player may not be impossible, but the implementation would be much more problematic than it would be in a less rigid system. Up to this point, my first priority has been ensuring enjoyable "standard" games. For the Something Awful Dominions community, "standard" tends to be 12-player maps with 15 provinces per player and no custom nations. I think that the current version succeeds at that goal, albeit in a pretty bare-bones state.
Next time we'll discuss the future, including a likely interface overhaul, possible graphical improvements, greater support for mod nations, and hopefully cross-platform builds.