Retro D&D and the old-school renaissance fascinated me, from a game design perspective. Although I'd say the movement reached critical mass about a year ago (largely fulfilling its need: bring the old versions of D&D to the OGL), there were some noticeable trends that were interesting.
• None of the popular OSR games were actually close-as-possible recreations of the existing games. Most were faithful in spirit, some were close, but none were actually as-close-as-questionably-legally-possible. Even my eventual favorite, Labyrinth Lord, took some liberties with the system it emulated. Some mechanics were adjusted, and some content that had no analogue in the original made it to the print run.
• Despite being tweaked and house-ruled, none of them went so far as to pretend that they were in fact a separate game, or an evolution or refinement of the game they were based on. This is a semantic argument I know, but when compared with the point above, it puts most retro-clones in a sort of middle-space that is neither genuine or inventive.
Of course, inventiveness and authenticity usually wasn't the stated goal; the aim of most OSR products was compatibility. You could write a Labyrinth Lord supplement and players would know that it would fit in their Red Box game. As if the rules themselves were never meant to be played, but could be if needed.
OSR games left the door open for me, as a person who creates content more than buys it, to venture away from the OSR in one of two directions:
1) Go back and find the edition of D&D most like the OSR products I enjoy, or
2) Make my own role-playing game that accomplishes my wants better than any aforementioned game.
Just playing an OSR game didn't really make the list (not that it had to, having reviewed many OSR and AD&D products I've decided that Labyrinth Lord is in fact my favorite published RPG).
I like the classes, and the monsters and spells, and the levels and experience charts. The only gripes I have with most D&D things are fundamental assumptions about settings, and how general mechanics are interpreted.
My ideal D&D-like game would look more like this:
1) Die size corresponds to function. Something like the d20 used for resolving combat and saving throws, d12 for tests of skill, and d10/d8/d6/d4 are used for hit points and damage.
2) Different character class paradigm. However, I have gripes. I don't like Clerics for theme, and I don't like Thieves for mechanics. Some of my favorite tropes are missing. I'd prefer concrete combination classes (like the Elf in Labyrinth Lord and most advanced classes in video games).
3) Tie character advancement rate on race instead of class, or perhaps both.
4) Get rid of the legacy fantasy mish-mash stuff that I don't intend to keep (like halflings in general, or PC gnomes). Additions should more fit common PC archetypes that are usually represented by bends in mechanics and logic (like magic-users that wield weapons, or vagabonds that aren't thieves, or spiritual warriors like berserkers, paladins and samurai).
5) Rework the core stats into a more reasonable split that can more concisely and accurately reflect various character mental/physical archetypes. There's no reason that Strength and Constitution should be separated. Intelligence and Wisdom are largely redundant. Dexterity lumps manual skill and speed together, Agility should be created to factor these two separate concepts apart. Things that consider whole strength-of-person (such as surviving resurrection, for example) might be divorced from stats entirely.
STR (Damage rolls, Hit Points)
AGI (Armor Class)
DEX (Attack Rolls, Warrior and Rogue Skills)
INT (Languages, Spell Charges, Saving Throws)
CHA (Reactions, Followers, Priest Skills)
d4 = Weakest hit point modifications (Priests and mages hp, soft weapon damage)
d6 = Normal hit point modifications (Rogue hp, environmental/magic damage, small weapon damage)
d8 = Strong hit point modifications (Warrior hp, monster hp, heavy weapons)
d10 = Supernormal hit point modifications (two-handed weapons)
d12 = Skill test (Rogue Skills, Priest Skills, Break Down Doors/Lift Gates, Listen, Initiative)
d20 = Attack rolls and saving throws)
Acrobat (AGI/DEX Skills related to mobility, personality and intrusion)
Avenger (Anti-Paladin, more evil, stronger power, no spells)
Berserker (Similar to 3e Barbarian, but with spiritual/religious servitude/compulsion/possession)
Fighter (Vanilla fighting class, as usual)
Ninja (Acrobat with some Fighter/Assassin/Monk capability)
Paladin (stronger power, no spells)
Priest (Like Magic User, uses Priest spell list)
Thief (AGI/DEX Skills related to stealing and intrusion)
Sage (Like Magic User, uses Druid spell list)
Sorceror (Like Magic User, uses Mage spells)
Wizard (Combination Fighter/Sorceror)
Some classes will be more powerful than others, and progression may be possible, e. g. Acrobat to Ninja, Fighter to Berserker, Sorceror to Wizard.
Human (any class)
Half-Elf (Acrobat, Avenger, Berserker, Fighter, Ninja, Priest, Thief, Sage, Sorceror, Wizard)
Half-Orc or Half-Ogre perhaps (Fighter, Berserker, Thief)
Light Elf (Acrobat, Fighter, Thief, Sage, Sorceror, Wizard)
Dark Elf (Acrobat, Avenger, Berserker, Fighter, Ninja, Sorceror, Wizard)
Dwarf (Fighter, Berserker, Thief)