Geromino: Nothing stops you for example from making say a D&D Wizard with the stats strength 16, dexterity 16, constitution 16, intelligence 8, wisdom 16, charisma 16, i.e. using their main stat, intelligence, as dump stat.
Actually, back in the pre-3e AD&D days, there actually was something; each class has ability score requirements, and if you don't meet them, you aren't allowed to choose that class. (This also meant that there's a small chance that you could end up rolling a character who doesn't qualify for *any* class, forcing a re-roll of the character, unless the DM is sadistic enough to force the player to play a commoner.)
Incidentally, I think the 3e-era rule (don't know if there's still something like it) that you need a minimum score in your casting stat to cast spells is bad design. particularly in a game where you cant really fix it after character creation (stat increases are few and far between, and do not keep up with the increasing stat requirements for higher level spells).
Some players, at least of the 3.x CRPGs, have fallen victim to this issue. For example, a player, not knowing the game mechanics, might create a Sorcerer with 16 Intelligence and only 8 Charisma, then find that the character isn't able to use any spells!
Edit: If I had to choose, I might prefer the AD&D approach of banning the character entirely rather than the 3.x approach of making the character non-viable. Making the character functional (that is, at least able to cast spells) but weak is what I'd probably go with, however. With that said, I prefer systems where you can fix those mistakes, like you typically can in SaGa games. (I can, for example, give some strategies for rescuing a party that's reached the end of Final Fantasy 2, but is having serious struggles due to utilizing the game's trap options, whereas there's no rescuing a D&D character with terrible stats.)
Geromino: You can call that bad design, sure, but its pretty hard to offer a player a ton of choices, as roleplaying games are known to do, and then design those choices in such a way that there are no poor choices possible.
I'm thinking that whether it's bad design depends on some things:
* Is it something a first-time player would reasonably expect to work? (Multiclassing is something I'd expect to be viable, even if not as good at any one role as a specialist would be.)
* If the player makes the wrong choice, is it feasible for the player to fix it later on?
If the answer to the first question is "Yes", but the answer to the second question is "No", then I'd say it's bad design.
In particular, Dungeon Siege, I think, fails both tests when it comes to multiclassing. Your D&D example passes the first test but not the second, so it's not necessarily bad design (though it perhaps is if we're looking at a Bard or Sorcerer). Final Fantasy 2 fails the first test, but passes the second, so it's not bad design (OK, maybe there is some bad design here, but at least the player can fix things if they know how things work. But, of course, that brings up the issue that games should be reasonably transparent in their mechanics; FF2 has some non-obvious (and, in one case, outright hidden) mechanics that should have been made more transparent to the player.)
