Close. You have the right idea, but your numbers are wrong. Strength (measured as power or force) is proportional to the cross-sectional area of the muscle (and only when you assume the number of sarcomeres per unit area is constant). Doubling a creature's size in all dimensions will increase mass by 8x, and muscular strength by 4x. (I studied Biomechanics in university).
So if you take a 4' tall halfling's strength as X & weight 65lbs., then a 6' tall comparable human would have strength 2.25X and weight ~220lbs.
My mistake then. I knew it was far short of the cubing of mass, but I guess I got it wrong about how much it did increase. I was fairly sure that it was different than bone strength, though, so is bone the one that only doubles?
Of course, this is only talking about direct muscle power, which is a bit different from effective strength because of all the other factors...
If the 3-18 bell curve is meant to represent human limits, why would the same bell curve represent halfling limits? Should the strongest human be no stronger than the strongest halfling? And if you don't penalize a halfling's strength, why do you give a bonus to strength for a goliath? How about a giant?
So I'm for bigger ability bonuses and penalties than they've ever done before. But that comes with the caveat that if they continue to make combat dependent on a single ability score, then I wouldn't want any bonus or penalty at all. Because at that point it's not really strength or dexterity or anything that the ability scores are actually measuring... it's class effectiveness.
I'm more of the opinion that the game should move away from the idea that ability scores, particularly strength, shouldn't be treated as absolute empirical measurements. They should be relative quantities that admit that they are rough abstractions designed to differentiate characters. In other words, it is not a problem to say that a halfling and a human both have the same general Strength Score range, even if the range of weight they can bench press is totally different. The reason for this is because Strength is an abstract quantity created for game purposes, not an absolute measure of a single, easily-identified physical characteristic. It exists to make distinctions between individual characters, not model physics.
Hmm, not sure if I'm explaining this correctly or just rambling, so maybe I'll explain my views a different way... Let's look at a different stat that makes this distinction a bit clearer. Dexterity is a good choice.
Dexterity is a vaguely defined abstraction. It folds lots of totally different qualities, such as manual dexterity, flexibility, balance, reaction speed, and so on. It is pretty much impossible to come up with some system that measures it in an empirical way, especially since many of those qualities have very little to do with each other. Saying that a small creature should be more dexterous, while a large creature should be less dexterous, only makes sense for some of those ideas, but not others. Ability scores don't measure real physical qualities, so it is fine if they don't increase and decrease directly with real physical qualities. A halfling and a giant having the same strength score is perfectly acceptable within the level of abstraction provided by the ability score system.
Assuming your numbers are correct ([MENTION=23654]dangerous jack[/MENTION] disagrees), why would the giant use a 8x as heavy sword if it can't swing it? It could swing a 2x human weight one as easily as a human his own sized. In fact, that's what large monsters/characters in 3e do - large weapons are only twice as heavy as medium ones.
If 4x strength is more accurate, that means the giant can swing his 2x weight sword twice as easily as a human his sword.
Well, this last statement you make isn't really true, for two reasons. First, the giant has to account for the increased mass of his own body and arm when swinging a sword, so it is impossible to swing "twice" as easily with a double-weight sword because his heavy arm would slow the sword down. Second, the change in leverage because of his longer arm further decreases the effective strength of his muscles (the human arm is a very inefficient lever, and gets less efficient the longer it is), dropping it even further.
Simply put, the idea of "strength" is an abstract quality with a huge number of physical principles all colliding within it. For every factor you take into account there are two more affecting the situation that you don't. That's why oversimplistic approaches like "things get stronger when they get bigger!" are not the best basis for game rules. It's best to just accept the abstraction for what it is and not think about it too much.
If you interpret ability scores as anything but the coarsest of abstractions, then even the idea of humans having a physical or mental ability range as wide as the 3-18 gulf becomes silly. Especially when a difference of merely 2 is the difference between human and halfling bench-pressing ability. If the total difference between humans and halfings is less than the difference between two random humans, then why do we need to differentiate races with stats at all?
I suppose it is a bit weird that I got involved in this whole physics-based discussion considering that is my opinion, but there it is.