When I first put together the idea for the Demonsword Project--the setting, not the game--I put it together in order to gather my thoughts about fantasy and magic so that, in my various stories and daydreams, I would have a consistent flavor for how magic worked. Magic in fantasy has been in many ways my obsession since I was a child; I spent months or years worth of time in my own head telling myself stories, because it was fun. I love magic, in principle; I love the idea that when you get the right idea in your head you can change the world to make it just so. More than the fantasy staple of magic as either set dressing or weaponry, I saw magic as a fundamental tool; a great magician could build, repair, manipulate, and in many other ways, tailor the world to his whims.
This, however, is a fundamentally dangerous and difficult task to take on for roleplaying games. The genre has been called at times "Collaborative storytelling;" setting rules for storytelling means that you must put numbers to a player's achievements, and while systems nowadays are fairly comfortable with conflict resolution--battles of steel or wit in particular--for the most part nobody is considering the math behind places, things, and magic itself.
Consider for example a group of players who have set themselves the task of defending a small town. An earth mage in the group could create towers for guards, or walls to keep bandits (or monsters or armies) out. How effective is this defense? How long does it take to create? How durable is it? If he wants to add, for example, parapets to the wall (that's the smaller wall along the outer edge, so people standing on the wall have cover), how good is he at that? What is the relative defensive advantage of the parapet? What is its durability? Does the attempt to create it weaken the existing wall? How does he know whether he has damaged the wall--does he have innate magical senses, must he study it deliberately, or does he have to "cast a spell" to get a sense of the wall's condition?
This task is fundamentally one that falls into a gap between skills, or alternately, it is a task that requires the merging of different skills. It takes equal parts knowledge of walls, familiarity with the building materials, and magical talent. Knowing wall construction and having magic power does not necessarily let you build a castle in a swamp, or a desert; the materials available may thwart you. Knowing the materials and having power also is not enough; you can throw together an earthen berm but any one of a dozen mistakes might be made in its construction: the walls could be coarse or sloped enough to be scaled, or bricks could be stacked and not held together, allowing them to topple in the face of assault; there might be no defense against ladders at all, and it may have no room for patrolmen to stand on it. The gate, if there is any, could be too wide to cover or too narrow for carts to pass, etc.
From a storytelling perspective, the game leader wants to be able to paint a picture with words given whatever information he has on the characters. An earth mage can build a wall, or a tower, but what interesting developments come as a result? Perhaps he needs the help of a local scholar, or a craftsman from a neighboring city, in order to make a decent effort at it. This picture must also be at least trivially functional; is the wall sufficient to keep out mundane bandits? What about an exceptional bandit? A well-equipped army? Or a squad of foreign brutes tasked with setting your town alight?
This combines with what I fundamentally want to do with the Demonsword Project, which is have a consistent, functional magic system. When an earth mage's berm is set upon by a fire mage, who wins? It may have been years in-game and many sessions since the wall was constructed. Is it fair to task the fire mage with bypassing the wall as though it were one made by thousands of gold and hundreds of slaves hauling stone for dozens of professional masons? Is it fair to give the earth mage's work a penalty compared to mundane works of mortar and stone? Is it fair to say it deserves a bonus, in spite of being weathered by years of normal usage? What if the attacker is not a fire mage, but water or wind? What if it is attacked by magic creatures? What if it is attacked by another earth mage--or by its original creator?
What most of this comes down to is establishing canon. Magic itself must have canonical parts: is earth magic superior to expert masonry, or inferior? Is fire magic harmless when faced with solid stone, or can it bore holes through feet of rock given time? But even with these questions answered, you must establish canon at the moment the skills are used. Is this a superior wall, or an inferior wall? Did he add parapets when he built it, or didn't he? How much cover do archers on the wall have? How easy is it to defend the gate? These questions need to be asked, and then answered, at the time that the wall is created, although the answers may remain hidden until they are relevant.
In this sense, I am interested in a faux-digital world like that of Erfworld; one in which places and things have parameters, and how well you do against them requires matching your skills against their level, exactly as you would match your skills against another opponent. But it goes beyond the obvious physical canon; if you create a magical shield bubble around yourself, how does it stand up to abuse? If you need to use that shield to protect against a cave-in or to stop a runaway carriage, can you? If you are capable of creating an impenetrable shield, can you create a flexible one, to catch a falling companion?
This is all why I have gone years without making much headway on the Demonsword Project RPG; it's discouraging to think that I have to account for a nearly infinite set of variables, and yet, I would not in good conscience release a project that I could not tell an interesting story in. Early prototypes had giant dicepools, and would tell you a fair amount about the result of your skill roll, but how do I apply that? I thought of adding effects you paid for by spending from the total die pool, but in the end, that simply adds far too much to consider.
So I'll keep looking, scratching my brain, and coming up with potential answers. It's fun, in a way. But it's also the most challenging thing I've ever seriously tried to do.