That D&D posits the role of an omnipotent GM is precisely what is at stake in the contrast between approaches. You take that view (which I described upthread as "unbridled GM force").
I find the idea that a DM could "cheat" or that rules could supersede his judgment equally controvrersial.
I think the role of the GM, in 4e and in classic D&D, is tightly constrained. (Eg in classic D&D a GM who introduces an extra complication to stop the PCs getting the gold through some clever stratagem is cheating.)
I am in the process of moving and don't have easy access to the books themselves, but even from memory, there are a number of passages that address the issue clearly in the 3e books (mostly the DMG and DMGII), including the "rule zero" text but also repeatedly throughout the basic descriptions of how skills work (i.e. the DM sets DCs and decides if and when rolls need to be made), the "DM's best friend" text (the DM should apply modifiers to any die roll based on his assessment of the circumstances), the DMGII bit about die rolling conventions and fudging (which notes that "DM's can't cheat").
References aside, I think it's pretty clear from the way the books are delineated as to what the DM is in charge of; i.e. there is a DMG that tells you how to referee and how the world outside the PCs works. I don't know what the "classic" DMGs have to say, but IIRC one of the most quoted passages from one of the earlier ones emphasizes that the rules exist to promote and enjoyable game and should be used only as appropriate to achieve that goal.
What you're suggesting (designer force, if you will), is that in at least some iterations of D&D, the text explicitly requires you to use the basic rules (hit points, saving throws, skills or their earlier iterations, whatever), even if those rules produce results that are nonsensical, unbalanced, or simply not fun. I'm skeptical of that approach. I think it's always been the case that the rules can and should be changed or ignored, and I think the DM has always clearly been in charge of adjudicating when that should happen.
As a separate but related issue, I don't believe there has ever been much in the way of narrative control mechanisms for players in any version of D&D. At the moment, we're talking about the DM's judgment versus hard rules, but the players don't have much of a say either way.
You can see this type of constraint to an even greater extent in (say) Marvel Heroic Roleplaying, where the GM has a finite though manipulable and growable resource (the Doom Pool) and must spend Doom Pool resources to do things like split the heroes ("The floor collapses under Cyclops and he drops out of sight!") or have Captain America's shield not return when thrown or bring a scene to an abrupt close.
Indeed. That is a game that clearly posits different roles for the participants. I'm mildly curious to try it out. It's quite a paradigm shift from D&D though.
Roleplaying is negotiated imagination. In order for any thing to be true in game, all the participants in the game (players and GMs, if you've even got such things) have to understand and assent to it. When you're roleplaying, what you're doing is a) suggesting things that might be true in the game and then b) negotiating with the other participants to determine whether they're actually true or not.
...
So look, you! Mechanics might model the stuff of the game world, that's another topic, but they don't exist to do so. They exist to ease and constrain real-world social negotiation between the players at the table. That's their sole and crucial function.
And they provide a framework for that negotiation
by modeling how the game world works. At least, in some cases; in a more narravivist/storygame approach, they might provide a different framework.
So you're sitting at the table and one player says, "[let's imagine that] an orc jumps out of the underbrush!"
What has to happen before the group agrees that, indeed, an orc jumps out of the underbrush?
Well, that depends on what the group wants. Many of those things fall under modeling the world. The group must agree that there could conceivably be an orc, and that it could conceivably be hidden there and be jumping out now. That's the world modeling function of the rules. D&D generally addresses these sorts of questions.
Then, there could also be a set of other considerations. Is it enjoyable and interesting to meet this orc? Is the orc a balanced or fair encounter at this moment? Does the orc fit thematically with the events going on at the moment? Do we have enough tome (in the real world) to deal with this issue before we have to leave? D&D generally doesn't explicitly address these issues.
An rpg could do so, but it's not necessary to do so. Certainly issues outside of modeling the game world ("balance", "fun", "immersin", etc.) are more abstract and more variable and more difficult to address with game rules. Whether this effort is worth the trouble and sacrifice is something for a game designer to try and decide. However, to my knowledge, D&D doesn't go there much. I don't know any version of D&D that gives players a say or not in whether the orc jumps out of the underbrush or outlines any sort of negotiation on that issue.
