In the DnD Next blog post this week Mike Mearls talked about classes and how people approach them. He discusses how some players have a concept in mind when they begin and they try to find a way to hammer a particular class into the concept they started with. Others window shop and when they see a class they want they just play it as it is. One group wants the system to present them with a mostly fully formed character they can just step into and the other wants the system to give them tons of options to tweak things just so. What ends up happening is the system sits partway between those two extremes, offering classes that have some fixed features and some options.
My group of friends built a system years ago that discovered the necessity of a middle ground the hard way. We designed the game down to three generic classes, then two, then one. A player could build any combination of thugging, magic use, skills, and tricks but people seemed very uninspired to play the game. Having restricted options and built in flavour really increased the fun of character building. People may say they want complete flexibility but it seems that few really do; note how few people actually play GURPS.
What is interesting to me is that Mearls tries to position DnD 3rd as the edition where people had tons of choices and 4th as being highly restrictive. I think it is true that you had more options in 3rd but the problem was that nearly all of those options were garbage. You could be a multiclass wizard / ranger / cleric / bard / rogue if you really wanted to but since you would be completely useless that doesn't really count as an option. I think 4th edition actually had significantly more options that were decent but it certainly cut off all kinds of junk builds by fiat rather than by attrition. In 4th edition as long as you maximized your attack stat you couldn't go that far wrong - in 3rd by comparison you could make a monumentally crappy character with ease.
I suspect that the final model for DnD Next's classes will be one that supports both philosophies reasonably but also generates maximum revenue. I would bet that we get core books that present a few simple options but mostly let you just choose a class and go and then supplements that provide plenty of ways to swap out baseline abilities for new and interesting things. People will complain of course about splatbooks and power creep and the necessity to buy tons of junk but that model lets new players buy a Player's Manual and get started easily and lets the hardcore people buy eighteen books and make their super optimized twinks. This is of course the model that was financially very successful in both 2nd and 3rd edition already.