You seem to be missing quite a number of factors; perhaps you've nevr
tried any of these things you're talking about?
* A person on their knees, or crawling, is not comparable to a very
short person for most purposes, including manual dexterity. The
relative proportions of different parts of the body are very
important.
* People who are short but quick *do* tend to excell at basketball, at
younger ages, until their taller, larger compatriots develop the
coordination necessary to interfere (such as blocking shots). Even
then, in the skewed world of Professional basketball, the skilled
shooters are almost always at the smaller end of the size range.
Most importantly, a professional *anything* will very likely have
skill modifiers (i.e. ranks) that dwarf the effects of size and even
talent (ability bonuses). In your example, this is even more likely
to be true, since it would be very rare for a halfling and a giant
to have the same dex.
* In the real world, the size of the *missile* versus the size of the
target is an important factor. D&D 3e decided to simplify this to
the size of the *missile wielder*, since that last already included
rules for weapon (and thus missile) fire.
* Real world analog of skills typically depend on much more than `1
stat', but D&D 3e makes some simplyifying assumptions for game
balance (and tries to compensate somewhat with synergy modifiers).
For example, I can assure you that when rock climbing, dexterity is
at least as important as strength. In fact, when actually leading a
pitch (what most people would think of as `climbing a cliff face',
for example), dexterity, constitution, and intelligence are all
likely to be more important than strength. A more `realistic' system
would use combinations of these, such as ``you include strength
penalties but not bonuses. You may apply dexterity modifiers of no
greater than +2. Any pitch longer than 25 feet requires either an
climbing+int check to find a safe resting point or a climbing+con
check to tough it out without resting.'' I personally encourage
everyone to use these sorts of systems when appropriate, but I feel
quite understanding when thinking about why WotC didn't include them
in the core rulebooks.
* your example of the `Giant-Halfling war of 794' totally ignored the
critical point that a group of 11 `typical' giant soldiers (say, 3rd
level) will, assuming mostly average rolls, totally destroy a group
of `typical' 3rd level halflings with shortbows whilst losing at
most 2 members of it's own group -- and that assumes that the giants
aren't being very smart (perhaps a reasonable assumption :-). Yes,
they may hit %20 less often, but that ignores range effects,
relative hitpoints, and relative damage from hits.
Your example `war' much more closely models a game of `Giants versus
Halflings paintball', where the goal is simply for each side to hit
the other with identical effects more often, but one team is made of
small, nimble, quick people and the other is made of larger,
clumsier people. This is, incidentally, why the reliable,
affordable, personal gun had such a *huge* effect on combat -- the
game shifted from `dealing damage' to `get any kind of hit at all'.
