The medium of tabletop RPGs is defined by its liminality. In practice, this means that players will freely move between what--to outsiders--looks like a series of connected minigames that share some common lobby. (Yes, that comparison to Mario Party is intentional.)
What they don't see is that participants, being in the moment, do not perceive this to be the case. Because tabletop RPG play is in that liminal space, play passes from (e.g.) exploring a dungeon to chasing a dragon to figuring out how to haul out its treasure to managing an expedition all without missing a beat. Alternatives to the medium succeed by carving out one part, focusing upon it, and mechanizing it to meet expected gameplay experience targets- things all done by the designer and publisher, not by users.
A proper tabletop RPG leaves those decisions to the users. What is the setting for your campaign? That's determined by the Game Master, not the publisher. What is the campaign going to focus its activity around? That's determined by the players, not by the designer or the publisher. As noted in the comments here and at Google Plus, the best thing that the designer and publisher can do is to give crystal clear guidance to the Game Master as to how to make his specific campaign the most entertaining game possible.
This means that the ideal proper tabletop RPG doesn't build up so much intellectual real estate that there's no room to maneuver without demolishing something. That's what happens with settings possessed of well-known and developed canons (including real history), and that's what happens with rulesets with too much mechanical complexity. In both cases, the liminality is destroyed.
The designer and the publisher must seek to stop at the point where that liminality is damaged, and it is better to prevent the issue altogether by making the use of game jargon something used minimally in favor of ordinary natural language. I've gone on about this at length previously, and you can review those posts here, so I won't belabor the point here.
You can learn to feel out where the weight of the rules constricts possibility of play instead of expands them. You can learn the same with setting weight. In both cases this is a learn-by-doing process, where you go over the rules or the setting and then see how they changed over time until they got too fat for their own good. Rulesets and settings, like the people who make them, have a healthy weight wherein they are at their best; too fat and they collapse under their own weight, and yet go too thin and you have no "there" there to justify it as a game.
That healthy weight is where the power of liminality is maximized. There's just enough substance to give context to emergent creativity, but not enough that it gets in its own way and thus constricts it. In setting discussion, this liminality is usually in the form of a sandbox; defined boundaries, and often a palette guiding flora and fauna (including monsters), but nothing else- especially no fixed continuity going forward or NPCs that do the real power-playing. In rules, we usually talk about ruling-friendly or GM-friendly rulesets where Mech Piloting is not a good idea (or even viable). Both are pracitcal expressions of liminality, and a commercial tabletop RPG maker has to embrace liminality to succeed in this small--but influential--gaming and entertainment niche.
That's why I call liminality the killer app; once you grok it, the knocks on the medium reveal themselves as strengths.