There's been a lot of talk lately about skills in D&D 4e. Sarah Darkmagic's last two posts (here and here) are a working summary of the current round of this discussion, and I can't say enough good about some of Rob Donoghue's posts on this topic, though you'll have to do some digging to find them at this point.
The point that hasn't quite been addressed yet is the inequality of demand between skills. This is an issue that is well-nigh universal to games that include a "normal" array of skills. (Technoir, with its tightly constrained list of verbs, doesn't have this issue. There are some things everyone needs to be able to do, and some things only one person in the party needs to be able to do. For example, if there's a chasm, probably everyone will have to cross it (possibly with some help from their friends). If the whole group wants to sneak from Point A to Point B, every character in the group needs to roll Stealth. If an opponent is hidden, everyone needs to roll Perception to spot the enemy. Let me clarify "need" in this case: parties can adapt to a small percentage of PCs being able to hit the target DCs, typically by doing something that lowers the DCs or increases the bonuses for the other PCs' rolls. There's something to be said for this kind of teamwork, and I'll come back to this point in a minute.
At the other extreme, the game has skills that need just one specialist per group. Conveniently, these skills generally represent highly specialized training. Arcana, Dungeoneering, Heal, History, Nature, Religion, and Thievery are all examples of this. Heal is the kind of skill that newbie players imagine that everyone in the party should have (by comparison, I believe just about every PC in a LARP should pick up some way to stabilize the wounded), but in the games I've played, it's really only useful to the party member with Ritual Casting - my players didn't have the patience to treat diseases nonmagically, if they even remembered that it is possible. The other functions of Heal are good, except that there's almost always a better way to accomplish the same thing using a power.
As a quick side note, Pendragon not only has this problem, it egregiously exacerbates it, as the main way to improve a skill is through use. If you aren't good enough at something to be the one the party calls upon to use it, you're never going to be, unless the party is split up.
Lore-oriented skills are subject to this kind of specialization in every game. D&D 2e and... well, a lot of games made in the 90s famously suffered from this, as they created monstrously long skill lists that no player in his right mind would ever think of learning. Massive skill lists offering pointless levels of specialization are the conceptual reversal of the (warning: TV Tropes link coming up! Do not click it if you have something else you have to do today!) Omnidisciplinary Scientist. The Omnidisciplinary Scientist is, well, also the Brain of the Five-Man Band, and thus is awfully useful as a niche in a PC team. D&D stats (and SIFRP stats, while I'm thinking about it) lend themselves to creating just such a character, as all Lore-type skills are based off of the same stat (Int in D&D, Knowledge in SIFRP). There's a very short list of reasons you'd have more than one character in a party pursue the same Lore-type skill:
They get the skill for free anyway (Arcana for several arcane-source classes, Religion for divine-source classes).
The players involved find it sufficiently valuable to have two chances to learn things about that skill, and they've got other bases covered well enough already.
The characters are satisfying prereqs for something else they plan to pick up (such as 3.x prestige class and feat prereqs), or the skill has other applications (Mage: the Awakening and rote skills).
The game in question has a combined-effort or Aid Another system that grants greater bonuses for hitting higher DCs and/or having more ranks in the skill (3.x D&D, Mage: the Awakening).
Miscellaneous roleplaying reasons.
What I'm getting at here is that some skills just aren't necessary for characters-in-general as others. Thievery is an interesting case here, as 4e is the first edition of the game since the creation of the thief in which one can (by hook or by crook) play a traps-and-locks rogue without explicitly being a rogue or thief (whatever we're calling it in the edition in question). 3.x let other character classes buy cross-class ranks in Open Lock, Search, and Disable Device, only to declare (through Trapfinding) that only rogues could even roll against some kinds of traps.
What all of this means is a little harder to pin down. I certainly don't recommend creating a situation in which every character in a group needs to specialize in the same Lore-type skill. At the other extreme, Athletics (and similar skills in previous editions of D&D) isn't really a protected niche for fighters or rogues, and shouldn't be. The things that a character does with Athletics (or Climb, Jump, and Swim) is generally bypassing physical obstacles in a straightforward way, and thus shouldn't be the exclusive province of one character over another. Acrobatics (Balance, Tumble) and Stealth, on the other hand, are flashy interaction with the world; the main things that characters do with acrobatic stunts in fiction are getting to hard-to-reach places and bypassing physical obstacles in an impressive way (the link is one of the more memorable displays of Athletics versus Acrobatics).
Just to round out my list, I'll mention that there are three more skills everyone in the party more or less needs. Perception - well, if you can't correctly perceive your character's surroundings, you can't really make informed decisions, now can you? Endurance is the kind of skill that either everyone needs or no one needs, depending on how much the DM likes diseases and long periods of physical exertion. Finally, Diplomacy (and to a much lesser extent Bluff and Intimidate) theoretically governs your ability to interact nonviolently with NPCs. One of the commenters in Sarah Darkmagic's second post on skills pointed out, though, that the game would not be diminished if every character received training in Diplomacy for free, and I think I agree.
I think the solution, at least for 4e, is to narrow the vasty seas between the best a character can be at a skill and the worst, and work on side benefits (such as the skill powers introduced in the PH 3) for characters who want to make that skill their area of expertise. A narrower range of skill scores would keep more characters in reach of the target DCs, and thus improve their ability to participate in skill challenges. The idea of simply giving characters more trained skills might work too, and isn't far from what I suggested some time back. Ultimately, I'm happy to quote Samhaine's point, that the skill system and skill challenges need a deeper integration into the game's mechanics.
There exists no finer example of a system that integrates skills with the rest of its mechanics than Technoir, but it achieves this in part by simply declaring that players do not roll dice unless they are struggling against an intelligent antagonist. Fortunately, struggles between man and nature are not genre-appropriate to noir... but I would love to see more done with them in fantasy adventure. The movie of Fellowship of the Ring does a lot with this - a lot of the physical heroics are outside of combat, such as Caradhras (though of course that has an intelligent antagonist, in the distance) and the crumbling stairs in Moria. The OSR guys are all about bringing the hexcrawl back, but 4e's skill system could make that kind of game really sing if it handled man-vs.-nature struggles with more grace.
To bring this ramble to some kind of cohesive conclusion, I think 4e pretends that all skills are equal, but doesn't create such a situation in actual use - not in availability, demand, or utility. I'd like to see this change, because I think that despite their faults, 4e's approach to skills was an important step in the right direction.