Optimizers are an important group, and shouldn't be marginalized because it can be taken to extremes. Role-players have the same issues in the other direction ("I don't want my Monk to be so weak, it doesn't fit my vision of the character, change something to make it work!"). Both of these problems can be fixed by better design.

However, when it comes down to it, I disapprove of the breadth and depth of the character creation mini-game in 3.5. It is the single biggest turn-off to new players, it is so much more determinant of how well you do than what you do while playing that it makes in-game decisions relatively meaningless, and fixing it so it maintains its robustness while addressing the former two is nigh impossible.

That being said, I am fine with compromising, so long as the core math of the game is, in fact, balanced. What do I mean by that? I mean that all classes, in fact that all builds for each class, can handle a roughly equal number of level-appropriate challenges, at least within a certain degree of deviation. So, let's say the difference between any two given builds in overall utility against all the challenges at that level should not exceed 20-25%. One Fighter build might be 15-20% better than another Fighter build, or another class build, but that's the most optimized build vs. the least, there. If that's the case, then your average party is assumed to not differentiate more than maybe 10-15%. Ideally, that would also indicate that while PC X is on average 10% below the party average, there are still a subset of challenges that he is the best equipped to handle.

If this is your balance goal, then it informs you of how to approach design; you create benchmarks for challenges (CR, mostly), and build your challenges first, with their differing immunities, specialties, etc. Then you design classes and feats and what-not which also match those benchmarks, with their own specialties, strengths, and weaknesses. Then you run the math; using average damage, high estimate and low estimate, and taking into account odds for SoDs and basic tactics, and you can figure out what percentage of CR X encounters level Y build A is capable of handling (hopefully 50 or above). That percentage is that particular build's score, and through rigorous testing you get to the point where no known build exceeds the designated variation in effectiveness from any other build. Ta-da, balance!