The other day I (respectfully) challenged B/X Blackrazor's JB's definition of niche protection vs. redundancy.
JB does a great job of explaining his idea of niche protection and redundancy here, so go read that first.
I think he's right on the money with niche protection in play.
However, I think D&D (especially the very oldest versions of it) has very strong inherent niche protection. Who turns undead the best? The cleric. Who can detect traps without burning up spells per day to do it? The thief. Who fights the best bar none? The fighter. And so on - and magic items restricted by class just further drive in that protection. Can anyone use this Staff of the Magi? Er, no, just the wizard.
So you don't need to worry about it too much in play or even in chargen. Even a 7th level magic-user will have stuff to do in a game with an 11th level MU with similar spells to select, because you never seem to have enough spells to go around. And they'll likely end up with different magic items, anyway. Where character classes duplicate each other, they usually have some niche aspect to it that either limits their options or isn't duplicated elsewhere. For example, Paladins have healing like clerics and fight like fighters, but at a cost of alignment, magic item, and equipment restrictions (and IIRC a steeper per-level XP cost). Assassins do the exact same things as thieves, but at a lower level equivalence, and get a special assassination ability that no one else gets. Illusionists in AD&D give up the "artillery" aspect of magic but gain in another focus of magic that magic-users don't do quite as well. All of this depends on well-made classes, of course - but the D&D classes generally balance well enough against each other in terms of choices and not causing total redundancy (see below).
Niche protection is a bit more of an issue in point-buy systems like GURPS because there aren't rigid class boundaries (or effective limiters like the level/xp costs of multi-classing). What's stopping your barbarian from picking pockets better than my thief is how we spend our points, more or less. There are a lot of ways around that issue (such as limiting point totals), some of which have to do with GMing and some with game design. It's a potential downside to the extreme flexibility of character generation. But it's not a non-issue for class-and-level systems. The more classes you have, the more concern you end up with about having too much redundancy. This is where the plethora of sub-classes of AD&D and overlapping kits and feats in 2e and 3e come in.
This is why, for example, GURPS Dungeon Fantasy has such strong niche protection for its templates - tightly constructed templates, lots of suggestions for reducing the options to duplicate another's specialty, restricted options for what you can start with on your character, and a smoothed path to upgrading what fits your niche. It's emulating a genre where strong niches are common and where allowing complete customization can weaken the emulation. In my opinion, anyway.
But in those point-buy systems, do JB's points hold up?
I think they do.
Like JB says, difference of play is key here. If your cleric can cast Find Traps and outdo the thief's trap finding, that's fine - who cares? It comes with a cost of a spell slot (or in GURPS, with learning See Secrets instead of something else) or of magical energy. It's not always on, it's not permanent, and it's not free. It's not always going to work. It's another way around the problem - JB's many ways to skin the cat argument. In a good game, sheer variety of circumstances will result in a need to have those multiple ways to skin the cat. Plus you'll often need multiple people doing something to defeat the problem/advance the plot/accomplish the goal. Not the least of which is if, as JB notes, someone flakes out/misses a session/gets his PC killed/etc.
But what does this show you? It's that redundancy isn't bad, either - it gives you more options.
What sucks is total redundancy. That's when your character (or worse, your character class/template/type) doesn't do anything better than the others. If your system's clerics get automatic, 100% reliable trap detection and removal ability, always on after a certain point, yeah, your thief is useless now except as a backup in his own area of supposed expertise. If the mage is a better swordsman than the swordsmen are, that's probably an issue. If you do A, B, and C and another character does A, B, and C+1, or A, B, C, and D, it better be because you're a less powerful version of the same character - less points or lower level, say. If not . . . you're stuck.
If it's because of lower level or less power or less attention from the GM, that's a play issue. If it's because your character class / template / whatever sucks, that's a game design issue. You can't catch up and you can't ever do as well as the other players who made a better choice than you. You may as well go stand in the back and hold a torch like the hirelings - make yourself useful.
I'll follow up tomorrow with a discussion of niche protection from my perspective as a game designer. Or at least as a writer of GURPS supplements.