Since it's going to take years to finish this game I have a plan to "start small" with exclusion of data for alpha releases. It means there will be only those features which are considered to be finished and the size of game world is reduced and so on. This could actually be one of the best ideas I've had for a long time, because in my opinion there are many good things about this and only few less good.
From development perspective it's easier to concentrate on finishing features when you don't have to work with a large amount of data. The data is mostly there, but it's not yet functional or fine tuned. The way it works mainly is excluding commands, because it's mainly commands which are also features and you handle the data through them. The second part is limiting the number of object types available. For example in the first release I'm planning to have only one playable character class so it gives an idea what it will be like.
This model also forces to finish those "simple" features which are often overlooked and left unfinished, because they are supposed to be done "fast" before the release.
The early release is probably going to be a positive experience mainly, because then it's possible to fix things that developers can't often figure out. Not only bugs, but some annoying features etc. I'm not worried about bugs, but I guess there will be some people who don't get it and they could complain why isn't the game finished yet. Also, I'm almost sure there will be "ideas" people, you know those annoying guys who throw in their ideas. They don't realize that developers are often creative enough to have their own ideas.
This plan is already in motion. The first part has been easy, because in data-driven engine it's quite easy to exclude features. Some features are excluded by commenting out source code, but it has been also quite easy. When the first release happens is not that easy to predict, but we'll see eventually how this idea works.