I've been pondering skill systems in RPGs lately. Most skill systems seem to work as follows:
1) A few skills (and/or attributes underlying those skills) are disproportionately more useful than others, so every player character tends to be built to emphasize those skills.
2) The skill system uses an escalating system of point costs (so that, say, going from 3 to 4 costs more than from 1 to 2), such that specialization is very costly.
3) The character development system does not require any maintenance of skills; characters can simultaneously maintain expertise in a wide variety of fields.
4) Any character can attempt any task, and reward from skill is largely linear. The difference between a skill level of 2 and 3 is +1, or 5%, or whatever the system uses. There is no qualitative difference between skill levels.
5) Characters acquiring new skills don't gain any benefit from already knowing similar, related skills. Even though you have Handgun +10 you somehow still have to start with Rifle at +0.
The net result is that characters tend to be mechanically quite similar. Everyone will tend to be competent in the handful of skills that are important to gameplay, with differences measured in a couple points here and there, with each character perhaps having one somewhat unusual skill ("I'm the Demolitions guy") but even there not being so advantaged that a good roll/Luck point/Bennie isn't more than enough to make up the difference.
The above outcome isn't necessarily a bad thing. For campaigns wit small numbers of player characters, role-playing super-competent heroes, it's great. James Bond does well with a skill system.
But for games that have large numbers of player characters, it can feel like no one PC really has a particular niche or uniqueness. The problem is especially bad in genres that don't have something like magic to spice it up (e.g. Westerns). Good role-players can certainly make any given bundle of numbers feel unique, of course. But a lot of players want more. I've had several failures in getting Traveller campaigns off the ground because my players wanted more meat in their character mechanics.
A lot of systems resolve the mechanical problems above by tacking on a system of advantages/disadvantages of some sort. Savage Worlds, for instance, has all attributes and skills confined to a tiny range (d4 to d12, basically 1-5) but manages to allow for lots of character diversity through its advantages and disadvantages.
But it seems like a totally different approach to a skill system might viably solve the problem in a different way. I imagine a system where:
1) The difficult in learning improving a skill isn't based just on how high your skill in X currently is, but in how high your skill in everything is - all skills would be totaled to create a "total level of skill" representing how much time and effort is required to stay good at what you already do. Kids learn faster than adults, etc.
2) To the extent that your skills are related your maintenance requirement would be lower - it's much easier to maintain and improve the set of skills Handgun +6, Rifle +6, SMG +6 then to be Handgun +6, Surgery +6, Arabic Language +6..
3) To avoid a situation where skills are just linear improvements, and create qualitative interest, each skill is associated with a selection of feats/advantages and increasing in skill allows you to choose those feats/advantages.
At which point I realize I've basically reverse-engineered a Class system with multi-classing.