This is an excellent essay on how to make the 3.x games more old-school in feel. I highly recommend it.
Article by "Matt Finch", creator of S&W. Posts as Mythmere.
I'm going to focus here on what "parts" of Old School play can be imported into the 3e rule system, and argue that 1e and 3e have lots of similarities (skipping an aberrational development circa the 2e period). I'll also argue that there's a limit to the 3e DM's ability to create an entirely old school game without being unfair to the players, but that there's a lot of capability to get close with the 3e system.
1) Definition of "Old School:" "Old school" is a term so broad as to be almost useless. It needs to be broken down into a couple of components. To my mind, these are: (a) presentation in the rulebooks, meaning art and writing style -- the least important component (b) relationship between the character, the player, and the Game World (c) nature of challenges (d) correspondence between rules and purpose.
2) Relationship between the character, player, and Game World. This is the biggest and most important distinction between the three eras of 1e, 2e, and 3e. In early 1e, and more so in Original D&D, the character was a playing piece at the beginning of the campaign. Characters had no backstory and adventured in a world that's unbelievably deadly by 2e standards and contained more chances for sudden death than 3e. I wouldn't say that the 3e framework (looking at encounter tables, modules, etc) is a whole lot safer than the 1e framework, although it is somewhat safer. The real distinction is that 1e had many threats which could just "kill" a character. Lethal poisons, save-or-die traps, diseases that PROBABLY would kill the character at levels before cure disease is readily available, etc. It's the volatility of the chance of death which distinguishes the risk level of 1e from that of 3e. This might seem arbitrary, but keep in mind that 1e characters weren't expected to take on their "life" until they had achievements. You carved out your history rather than growing into a backstory.
History of the Development of this "Character as Pawn" concept. 1e and 3e are somewhat similar in their approach to the idea that a character earns his history and doesn't start with much of one. I'm pointing to the way the rulebooks describe the "flavor" of the games. Obviously the DM affects this CONSIDERABLY. In 3e, there's a lot more work involved in creating the "pawn," because character creation is an area of the game where player skill is required. The 1e "pawn" was far less complex and far more disposable in the player's mind even than in 3e. You didn't really expect your first character to make it to second level on the first shot. After third level, you're getting attached to the character and you have a REAL sense of pride that he's survived. Between 1e and 3e, however, there was a decade of a different theory of the game. It started in the late 1e period, but blossomed with 2e (and was rife in the 2e rulebooks). This theory will sound alien to most 3e players, but the idea was that the character was NOT SUPPOSED TO DIE UNLESS ABSOLUTELY NECESSARY. Indeed, in Dragonlance modules, the "plot" could be derailed by the death of characters who played a role in the story. The focus was on roleplaying far over and above the live-or-die focus of 1e and 3e. Worlds needed to be rich in detail, for exploring these fascinating places tended to replace the brutal us-against-them attitude of 1e/3e, whose players expect combat at any moment. DMs who started gaming in the 2e period often have traces of this conventional sidom embedded in their methods; the idea was promulgated in Dragon, in the rulebooks, and everywhere. This generation of DMs was taught heavy-duty roleplaying and player-backstory rather than "carve out your destiny from nothing in a world that operates without your help or interaction." You could call the 2e method old-school or the 1e/3e method old school -- depends when you started playing, in my experience with DM conversations.
3) Nature of Challenges. As mentioned before, 1e contained no abilities that just turned on and functioned (like spot checks or intimidate checks). You had to describe what you looked at, touched, etc. Traps and clues were geared toward this method. 3e thus puts little emphasis on thinking through non-combat challenges. The 3e focus is more on combat tactics (much richer than 1e) and on parsing out the overall situation. This is an area where it would be very hard for a 3e game to simulate 1e without screwing the players. Your rogue player has to spend part of his initial resources on a limited number of chosen skills. If you de-emphasize spot checks, you are skewing a major part of the system in 3e. It can be done, but the 3e system is so integrated that changing one area tends to have unexpected results elsewhere.
Another point about challenges. In old-school play, it is a matter of player skill to choose what level of risk/reward they'll attempt. The best example is the dungeon level. Want to shoot for the moon and go down to level three when you're still first level, hoping for the big score? Go for it. This is also the case in 3e, although the written rules tend to create less variation in the levels of risk. In 1e, you could find some way-bad risks on that second level. In 3e you've got a pretty good idea of the risks if the DM does it by the book. Less chance of a total "we had no chance" encounter. The 3e rules need virtually no tweaking to get this part of the old school feel. Just put in a few lower and higher EL encounters within any given EL area.
In general, 1e players expected to face some risks that could just blow them away. Running away was an important part of planning. Know where you can use dungeon terrain to slow pursuers. Remember where you killed some of the orcs, and head back there with your next characters to fight the smaller number for the big treasure, getting your last character's scale mail as a bonus part of the treasure. 3e works like this, but less so. 3e players might even object to finding killer monsters because it's not balanced. The way to old-school your 3e game is to (as mentioned) vary the danger within a particular EL area, and explain to the players that reconnaissance has just become really important.
4) Relationship between rules and purpose. A 3e game can be made much more old school, but the rules themselves (spot checks, etc) can't be too de-emphasized without messing up the skill involved in character generation. Risk levels can be done easily, search checks can be made to depend on describing exactly what you do, etc.
5) Other stuff. Older editions were very archetypal. Your mage was obviously a mage and just didn't cross-train to give himself optimal respond-to-anything skills. He couldn't. He had to work with what he had, as a big factor in player skill. Good or bad is your decision, but there's a lot to be said for having constraints on your character generation. This part has a hugely different feel between 1e and 3e. Your skill in responding to threats outside your archetype skills was focused on changing your environment, not developing your character sheet. That vulnerable wizard hires some spearmen to surround him closely. Hirelings are a problem in simulating old school, btw. Putting ten spearmen into a 3e combat slows things down a lot. The more abstract 1e system allows much larger combats without adding much slowdown. I think a really good 3e DM could handle simulating this part of old school, but you'd need lightning speed resolution to keep mass combats moving quickly enough.
MY GUIDE TO MAKING AN OLD SCHOOL 3E CAMPAIGN
1) Restrict classes to the ones in the core books and eliminate the prestige classes.
2) Don't allow purchase of magic items
3) De-emphasize the experience gained for combat and offset it with experience for getting gold. Why? Because giving xp only for killing stuff means you don't win by eluding combat in a creative way. The reward system for players, the incentives, become broader and promote a wider range of "solutions" to things.
4) CONSIDER giving out xp bonuses for use of thieving skills, great combat tactics, etc. This in and of itself is NOT AT ALL like 1e, and in the hands of an average DM will backfire badly. What it achieves, though, is the result that characters level at different times, not all together. Sometimes one character gets to be the star, sometimes another. This is so big a risk for so small an old-school effect that I don't recommend it. However, that "I'm the star until you level up" was a cool part of 1e gaming.
5) Suck up their cash (they'll have extra since they can't buy magic items) with training costs, cost of riotous living, etc. 1e created a system in which you lived really high for a while, then faced abject poverty until you went out again. It felt adventurous - you've sort of got to see that in operation, I admit.
6) Cut down the number of cleric spells available. You might even forbid trading out your spells for cure spells ... I kind of like that development, though. Maybe limit it to one spell that can be traded out...I don't know. Something to keep the cleric's firepower in check. Cut out the offensive cleric spells.
7) Don't allow wizards to buy spells except for outrageous cash (suck up the cash). Wizards should have fairly limited repertoires and be forced to use them creatively rather than always having the exactly right spell for the occasion available.
Cool Look at Wilderlands of High Fantasy as the campaign if you don't want to make your own. I've seen it, it's very sword-and-sorcery.
9) Make them explore hex by hex to find places, making a map as they go.
10) Require dungeon mapping, and make them tell you right-turn, left-turn, etc. to get out of the dungeon. Make this worthwhile with teleports and lots of things that can misdirect. This heightens the sense of exploring. Yes, it slows things down.
11) Remember that there are doors that they just can't open, things they just can't identify, magic beyond the whisper of a pattern represented by the text. MOST of magic is beyond human kind.
12) slow down level advancement a bit, so that there are more combats and experiences between levels. They'll feel like they earned it and have more accomplishments under their belts by the time they reach high levels. In the current system, you can be a baron after a much smaller series of challenges than in old-school gaming. It heightens the sense of accomplishment.
13) Screw realism, screw ecology, screw explanations, screw economies, screw physics. The explanation is out there for why an ogre is wandering the city without molesting anyone until he sees the party. The explanation isn't what the game's about. Killing an ogre in a cool city brawl is what the game's about. If they ask why, tell them to figure it out. They may try. Their line of inquiriy will give you good ideas.
14) Keep everything local. Avoid planet-spanning evil and planet-spanning organizations of do-gooders. Avoid making magic into a substitute for technology. Great wizards might occasionally communicate through crystal balls and such, but it has no effect on the world. They don't hire themselves out to barons for use as a telephone. Since they're not part of a worldwide force for good, they don't even pick up the wiz-o-phone for their own purposes. Eberron is thus sort of the opposite of old school, and Forgotten Realms is pretty close. If you use the Realms, take out the Harpers.
15) Don't put the characters in constant or reliable contact with super-NPCs like Elminster or Bigby. Don't set up a situation where that NPC might ever, ever, ever, pull the character's bacon out of the fire. It's not a game of saving bacon; it's a game of keeping your bacon ... um ... raw, I guess. That analogy went to hell fast. I would eliminate Elminster from the Realms. Even Bigby only has local power -- some troops, a dragon, etc.
Finally, some sword and sorcery tips about the game world's flavor. You've already changed it significantly by not allowing purchase of magic items, potions, scrolls. Quadruple the cost in gold of creating these, too. Don't put your characters in the business of being manufacturers - it's an adventure game, for cryin' out loud.
Use place names from Arabia, Turkey, and Greece, mixed with a few totally inapposite European names like "Hrothgar" or "Lord Melchik." Use tons of traps and don't give experience for beating them. I said keep things local - and people have no idea what's within 10 miles of their villages. Oh - forget about a G rating. There are prostitutes and slaves, and the good characters shouldn't be incentivized by their alignments to attack "legal" slavers. Galley slaves and slave markets somehow add a lot to the atmosphere; plus, it lets you capture the characters if they'd otherwise die, if you want to give them that loophole once in a while.