I'd like to create a sprite-based 2D rpg, also heavily influenced by Dwarf Fortress, where terrain, cities, forests, rivers, people, monsters, quests, dialogue, groups, and stories are all generated procedurally. Push a button and an interesting game comes out. That'd be cool.
Well, Dwarf Fortress's scope is... impossible to match. I struggle to think of any game with even similar complexity. Making a Dwarf Fortress clone would take many years, however culling down the feature set (and the depth of each features) to the actual "fun bits" at least makes it possible. So it's not even cutting the breadth of features, but also their depth. No need for complex climate or economy simulations, etc. No need for insane descriptiveness of objects (save for artifacts). It's a lot of complexity that reduces its appeal to a lot of folks (though caters very well to a specific targeted audience). Dwarf Fortress with a touch interface and cute graphics won't magically broaden it's appeal -- the core game needs to be streamlined and no longer be daunting to newcomers.
For example, no massive world at worldgen with full history and legends. Most people don't want to deal with complexity, they just want an embark site to start in and get to work on. It takes a lot of time to create those things that the majority of folks would never notice or appreciate.
There's a lot of satisfaction with gathering/growing food, building basic shops and homes, defending against threats, creating a trade industry, and a bit of dungeoneering. However one of Dwarf Fortress's strength is it presents a lot of competing interests -- where you need to weigh building defenses with industry, etc. Plus the occasional "oh shit" moment where the game attempts to stomp on your sand castle.
From reading about DF, and especially the NYT profile of DF and the creators, I get the strong impression that DF is very much being done the hard way, even ignoring the C++ aspects, and one could cut down the complexity tremendously with good abstractions and things like DSLs.
(Think the Viewpoints OS strategy - with good enough abstractions and DSLs, you can do an OS or languages in very small sourcebases; I'm still in awe of their TCP/IP strategy - parsing the RFC illustrations! http://www.moserware.com/2008/04/towards-moores-law-software... )
