For purposes of this post, by "old school D&D" we mean:
1. Chainmail, 1971 (not D&D obviously, but its combat mechanics heavily influenced what came later).
2. The "three little brown books", 1974.
3. Greyhawk (or actually the three little brown books plus Greyhawk), 1975.
4. Holmes Basic, 1977.
5. 1e AD&D, 1978.
6. Moldvay/Cook, 1981.
7. 2e AD&D, 1989 (I don't want to argue as to whether 2e is "really" old school or not. For purposes of this post, it is).
Let's look at how each deals with three basic issues:
1. Basic Armor Effectiveness
Each edition assumes that there is a continuum of armor "classes" and that one may make oneself harder to kill by wearing armor of a higher (or lower, depending on how the terminology is used) class. It should be noted, however, that in the Man to Man Melee Table in Chainmail (p. 41), there are exceptions to this for certain weapons. For example, while, say, a dagger has a 72% chance of killing an unarmored opponent vs. only a 3% chance of killing an opponent wearing plate armor, a two-handed sword has an unchanged 72% chance to kill whether the opponent is unarmored or wearing plate, and a mace actually has an increased chance to kill the armored opponent-42% against unarmored and 58% against plate. On average, however, armor is helpful.
Originally, each class didn't merely represent a place on the continuum but represented a specific armor type or combination of armor types, or at least it did for armor-wearing men and some armor-wearing humanoids. Thus Armor Class 7 simply was leather armor, Armor Class 4 was mail (or "chain mail") with a shield, and so on. AD&D changed this, adding more armor types and slightly stretching the continuum. The classes now ranged from 10 to 2 (as opposed to 9 to 2) and, with the exception of a few classes at each end of the scale, each now represented a selection of armor combinations. So, Armor Class 7 was now leather or padded armor + Shield or studded leather or ringmail. Armor Class 4 was chainmail + shield or splint mail or banded mail. And so on.
2. Variable Damage by Weapon
In Chainmail, the 3LBB's, Holmes and the introductory weapon rules of Moldvay/Cook, all weapons do the same damage-1-6 points (or for Chainmail, simply, a "kill"). Though, notoriously for Holmes, if calculating damage per turn as opposed to per strike, the expected damage per turn of weapons ranges from 2-12 to half of 1-6. (Small weapons-daggers, etc.-strike twice per turn and heavy weapons-halberds, etc.-only strike once every other turn. Interestingly, this yields results for relative weapon effectiveness that are almost the reverse of the other editions.) Holmes also implicitly states that his combat rules are somewhat simplified and explicitly directs those desiring more detail to consult AD&D. Greyhawk and AD&D have variable weapon damages ranging from 1-2 to 3-18. For the optional or advanced weapon rules in Moldvay/Cook there is a smaller spread-1-4 to 1-10.
3. Variable Weapon Effectiveness vs. Different Armor Types
The idea here is that while bettering one's armor class is on average a good thing, it may be relatively more (or less) helpful depending on precisely which weapon is being used against it. Or conversely, some weapons will be relatively better than others at attacking opponents wearing certain specific types of armor. One standard example would be that low impact slashing weapons are relatively less effective against plate armor than, say, blunt weapons. So, an AD&D longsword might on average be better than an AD&D hammer (because it does more damage) but it's not as much better (and depending on certain assumptions might even be slightly worse) against a target wearing plate armor.
On the chronologically ordered list given above, every other edition-Chainmail, Greyhawk, 1e and 2e-has some table or method of simulating this.
I would argue that while having a rules mechanic that simulates variable weapon damage versus different armor types is laudable and, all things being equal, desirable, each attempt in old school D&D was a failure. This claim is not really controversial (though no doubt some would take issue with it). But looking at why each attempt was a failure may give us clues as to whether or not such an effort could succeed.
More in Part II...