If the presence (or absence) of the secret door is recorded and independent of either the GM's or the player's desires for it to be there at the time and the GM isn't goading the players one way or the other, how can its presence (or absence) possibly limit the choices of the players?
I'm not sure if you intend this as a rhetorical question or not. I will treat it as non-rhetorical, and answer it. To the extent that you intended it rhetorically, you'll probably think my answer inadequate - sometimes that happens in discussions among human beings!
Here's the sort of thing I have in mind - it's a bit underdescribed but hopefully clear enough to get us on the same page in respect of it:
<the prior events of play, together with GM narration, establish (i) that the PCs are in a stone building facing some bare walls, (ii) thay the PCs are being pursued through the building, and (iii) leave it open what might be behind the walls in question>
Player: "There might be a secret door that we could escape through in one of those bare walls - I seach for signs of one."
GM: "Make a [Perception, Search, Architecture, as appropriate to system] check."
<player makes check>
<GM consults notes, notes that the notes describe these walls as nothing more than plain walls with no secret doors in them>
GM: "You don't find any signs of secret doors."
Here's why I characterise this as a railroad.
We don't know exactly how it has come about that the PCs are being pursued through the buiilding - most typically, I think, that would be a consequence resulting from some recent bit of past play. But in any event, that pursuit is now the salient pressure on the PCs (and hence the players) in the game. In response to that pressue, this particular player has expressed an interest in the fiction developing in a certain direction - namely, that his/her PC finds signs of a secret door in the bare wall, so that the PCs might escape though it.
Now, because we're playing a game with "moves" and dice and stuff, rather than just round-robin storytelling, the player's desire about the fiction doesn't happen automatically. Rather, the player declares an action for his/her PC that folows from that desire. Success in that action declaration will meant that the player gets want s/he wants vis-a-vis the fiction (ie the PC finds signs of a secret door); failure means s/he won't.
(Note that this action declaration satisfies other typical constraints on player-side moves in a FRPG. For instnace, it is declared from the first-person perspective of the PC. And it's a well-establshed trope of fantasy gaming that bare stone walls can in fact have secret doors in them.)
In the example I've given, the player's action declaration does not succeed, but not because s/he rolled too low on the dice. (This contrasts with a failed attack roll.) It fails because the GM has
already decicded that it can't succeed.
That is the limit on the players' choice - his/her choice to have escape occur by way of secret door has been vetoed by the GM, by application of the prior worldbuilding/setting authorship. That is why I call it a railroad.
For completeness, I've written a further comment; because it's tangential, I've sblocked it.
[sblock]There is a completley different style of play, which is not about
salient pressure on the PCs but is about
puzzling out the maze designed by the GM, so as to extract loot from it. The OP distinguished the two styles of play; the latter is associated with classic D&D dungeoncrawling,. I don't think the notion of
railroading has any work at all to do in describing that sort of RPGing, because the game isn't about developing a ficiton at all; that's just a side effect of the players declaring moves that allow them to map out the dungeon and locate and recover the treasures. In this style of play there can be well- and poorly-desigend dungeons (eg too linear, or too many pit traps, or too little treasure relative to difficulty); and fair and unfair refereeing (eg rules that the monsters always find the PCs while resting, no matter what precautions the players take to reduce the risk of being found). But no railroading [/sblock]
If the check succeeds, you make up stuff related to the goal. If the players are making up details for the results of their own rolls, you are not needed as DM. If they are not, they are declaring actions in order to get you to make up stuff.
You seem to have missed the way that "say 'yes' or roll the dice" actually works.
If the GM calls for a check and the check succeeds, then the player's intent is realised, and so the only work the GM did was to contribute to the framing, to call for a check in response to the action declaration, and to set the DC. It is
the player's desire for the fiction that comes to pass (just the same as in combat: a successful disarm roll, for instance, isn't just a cue to the GM to make something up: it establishes a definite outcome in the fiction, namely, that the foe is disarmed).
If the check fails, the player's intent is not realised, and rather the GM narrates some consequence which, ideally (ie from the pont of view of a satisfactory aesthetic experience), was implicit in the framing of the situation In this latter case, the GM does all the work done for a successful check, plus has to establish and narrate the consequence of failure.
The role of the GM is therefore pretty clear, I think.
The Story Now style involves the players setting goals(rails) and the DM doing everything in hos power to keeping the PCs on those rails by making everything in some way important to those goals.
You are, I think, the only person I've ever met who thinks that saying "yes" to someone's request is railroading them!
But in any event, what you say is not accurate, because it ignores the narration of consequences for failure. These obviously should have thematic/dramatic significance, but will constitute
obstacles to the PCs realising their goals. Likewise the framing of situations that don't follow from failure, but have the goal of provoking dramatic/thematic choices by the players for their PCs: these will be
obstacles to the PCs' goals.