Falling: Assess falling damage at 1d6 per 10 feet fallen (linearly). This assumes a fall onto earth or wood; decrease damage for yielding surfaces (water, snow, mud), and increase damage for very hard ones (stone, metal, etc.)
Falling has an extremely weird pedigree in D&D, and I could write at length just on that (in fact, here it comes...). Consider OD&D – Where are the falling rules located? Only in the naval combat section (for being pushed off the deck of a ship; also in passing in Vol. III, p. 5 *). And what is the rule? 1d6 damage per 10' fallen – but with a saving throw, generating only a 1-in-6 chance per 10' of taking any damage at all. For example, a 20' fall has just a 2-in-6 chance to fail the save (4-in-6 success) - and thus, two-thirds of the time, will deliver no damage whatsoever!
The subject of falling is mentioned in AD&D's PHB in only the most cursory fashion: “It is probable that your referee will simply use a hit points damage computation based on 1d6 for each 10' of distance fallen to a maximum of 20d6...” (p. 105) A pair of Dragon magazine articles later assert that this damage should be assessed cumulatively (i.e., 10'=1d6, 20'=3d6, 30'=6d6, etc.) – this being a very short blurb by Gygax in issue #69 (as part of his thief-acrobat presentation, later reprinted in Unearthed Arcana), and then a full-page article by Frank Mentzer in issue #70 (asserting a Gygax claim that the original PHB language was a typographical error; however, this does not agree with other Gygax works such as module G2).
A somewhat later issue of the Dragon (#88) had what I consider to be one of the most inspired and challenging presentations for that era. That issue carried the article “Physics and falling damage” by Arn Ashleigh Parker, wherein a “proper” falling damage system was deduced from rigorous consultation to physics formulas, gravity constants, advanced algebra, wind speed, and reference to texts on skydiving (the result being quasi-similar to the original 1d6/10', with more damage assessed earlier on). And this article came as part of a debate, with reference to its own rebuttal article in the same issue by Steven Winter, “Kinetic energy is the key”, which used other physics concepts to argue precisely for the original linearly-assessed 1d6/10' rule. Imagine that happening in Dragon today!
In some circles, the 1d6-per-10' rule was commonly ridiculed (ignoring the cumulative revision) as permitting high-level fighters to leap off 100' cliffs without much fearing for their lives. There is some reason to this, in that fighter hit dice swelled up to d10, and Constitution bonus up to +4/die, while falling damage remained the same 1d6 per level over time. However, I'm convinced that the reaction to this produced one of the most atrocious disfigurements of the system in 3E: the “Massive Damage” rule, whereby a save-vs-death was called for when any damage amount hit the magic number of 50. Most players are under the impression that this was an optional variant rule, but in 3E, it was not.
For many years I was using Gygax's cumulative system for falling damage, feeling indeed that falls should be more perilous, and I also applied the idea to other environmental factors (such as heat, cold, thirst, and starvation). In general, I felt that if higher-level hit points represented less physical stamina and more “dodging/fortune-type” factors, then they should be devalued in the context of an unavoidable fall. However, two problems with that have occurred to me recently. First, there seems little justification that someone able to dodge a monstrous blow could not also be able to roll/spring/cover their head properly to avoid the worst effects of a great fall (or simply land in a lucky spot). Second, when I looked in the DMG to compute exactly what percentage of hit points were “fortune” at any level, I was dismayed to find (on p. 82) that Gygax had stipulated a system wherein the raw physical hit points grew at precisely a constant rate every level (before abruptly ending at level 7). If this were the case, then even under my former assumption, falling damage should increase only linearly through level 7 (at least).
So now, here I am back today, opting to assess damage at the old standard of 1d6 per 10' fallen. (Perhaps if I were playing 3E I would increase it to a base 1d10 per 10'.) I'm not going to use the save from OD&D because (a) it's simply an unnecessary complication, and (b) it makes too many 20' or 30' falls entirely without injury, which really is silly. It's probably a reasonable amount of damage if most hit dice are still d6 themselves (Fighters d8, maximum Con bonus +2), and surfaces like hard, jagged stone can boost this by +1 or +2 points per die. Other environmental factors will probably also be assessed linearly from now on. We shall see.
* Edit 11/23/11: I just learned of another place in OD&D that I had overlooked: in the Aerial Combat section. "Crash -- for every 1" of height a rider must throw one six-sided die for damage occurring from the crash, i.e. a crash from 12" means twelve dice must be rolled and their total scored as points of damage incurred by the creature's rider." [Vol-3, p. 27] Thanks, Grognardia!