Classes
For multi-functional classes like druids, any competent player should be able to play in any of the class's roles, regardless of gear or spec. Sure, a healing-spec paladin shouldn't be quite as good at tanking as a protection-spec pally, but he or she should be capable.
Every class should have some breadth of scope. In World of Warcraft, mages are just about pure DPS class, in part because crowd control has become almost unimportant. Why not give them a little tanking ability? Sure, they shouldn't be as good at it as warriors or paladins, but why can't they at least give it a try now and again?
In counterpoint to that, the specialized classes should always be a little better at their specialty than the mulitfunction classes. This is the tradeoff that keeps specialized classes valuable: any druid should be a decent tank (provided that the player knows how to play), but a warrior should be a slightly better tank, given equal gear.
There are two challenges here. First, we have to maintain balance between the classes, so that all classes are interesting: no class may be obviously superior to any other. This depends on number-crunching, playtesting, experience, and a willingness to adjust the classes based on real-world results.
Second, we need meaningful distinctions between classes. The most interesting way to do that is to give the classes different mechanics. There are plenty of ways to inflict damage (melee, ranged, pets, fire magic, frost magic, DoT forms of all of the above, etc.), surely with some creativity we can find new ways of filling the Control and Support roles.
For example, the classic way to be a tank is to soak up lots of damage and maintain aggro on lots of mobs with melee attacks and AoE taunts. This is fun, but it's not the only way to accomplish the goal (keeping mobs from killing the squishies). For example, one could have a evasion-based class, like a rogue, who focuses on not getting hit rather than absorbing a large fraction of the damage. More creative still, you could have a magic-using tank class that is very squishy, but can root, confuse, charm, or debuff enemies to prevent them from doing much damage. On a smaller scale, a sword-and-shield tank and a spear tank should have a slightly different mix of abilities, so a player can choose what fits their style best, and have a variety of different experiences even if they stick to one broad role.
Mage-type characters can use the classic elements of fire, frost, and lightening. But how about other interesting choices like earth magic, illusions, phantasms (messing with people's heads from the inside), or time magic to anticipate a strike and increase a sorcerer's melee effectiveness?
Miscellaneous Mechanics
Gloom
Lord of the Rings Online has a cool mechanic: big bads like Shelob, the Watcher in the Water, or Ringwraiths cause "Gloom," which penalizes players in combat, reflecting paralyzing fear. Even trolls may inflict some. Experienced characters can earn Courage, which provides a small aura benefit but mostly is useful to counteract Gloom. In this way, characters must earn the ability to stand against great evil on an even footing. Twinks can come along on an adventure, but they'll be handicapped. It works seemlessly, with a very plausible in-universe explanation.
Creature Play
Creature play provides a change of pace, especially for PvP-oriented players. Ideally, they should be as unlimited as normal species, with tiers, equipment, factions, etc.
Reputation
Reputation should be regional, with a degree of spread. You may become a local hero, but remain unknown in the city, but then again with many great deeds your reputation may precede you. As with other elements of society, reputation should spread along trade routes.
Rep could be class-specific, meaning classes of society not character classes. A noble warrior may be well-known among highborn but draw a blank in a country inn, while a ranger who stands up for the peasants along the frontier may have the opposite. A master thief will be well-known far and wide in the underworld, though the populace as a whole doesn't know of her.
Exploration
There should be more cities and towns. They add a great deal of flavor to a game, allowing each region to have its own style. They're also helpful in providing multiple start locations. Durotar is cool and all, but having run it a dozen times, I'd love to start in a troll village in the jungle or something.
Travel
To make travel faster and less boring, provide automatic transportation along roads. Yeah, it's purdy to watch the world go by from the back of a gryphon, but sometimes you just want to get from here to there. There could be a minimum tier to travel through dangerous areas.
Crowd Control
Crowd control abilities tend to be very frustrating in PvP. Stun lock is maddening, and being rooted, slowed, blinded, or otherwise debuffed isn't much better. When designing control abilities, we should try to give them less frustrating effects against PCs. For example, maybe a character caught in an illusionary prison should be able to escape more quickly with practice.
Character Development
Tiers
What's the single most established characteristic game mechanic of roleplaying games? No, not the hitpoint, that's almost universal to all games. Roleplaying games in particular revolve around experience points. So let's do away with them. Yes, let's dump this long-treasured measure of success. And with it, let's dump levels, too. Do I have your attention? We replace 50, 60, 70 or more arbitrary levels with a handful of tiers, maybe six or seven. Let's name them rather than using numbers, to encourage players to think of them as part of their character rather than mere milestones to the end. Here's what it would give us.
Without XP, there's no grinding, and no delicate XP curve that can never please everyone. Players that are into exploration can spend all the time they like in each tier. Players that want to play a specific dungeon, or join their friends that are at a certain stage, can jump up very quickly.
Entering a new tier would unlock the ability to develop additional abilities, one of which would be granted outright as part of the quest to unlock the tier. Each specific ability would be unlocked by completing a quest of some sort. This means that a player could pursue whatever they find most interesting: want to go learn the next bigger and better fireball? Go for it! Would you rather travel to far lands to follow rumors of a mind-control spell taught by a secretive clan? Rock on!
Characters would need to develop their skill level in each abilities, so at first they may be unreliable or mana-intensive. This allows for the interesting situation where a highly practiced character goes up against a high-powered opponent with flashy skills.
If two characters in the same tier fight, the outcome will be determined by skill, equipment, and luck. If we downplay equipment, skill becomes the most important factor. A character from a lower tier could be a threat to a higher-tier character; although the lower character would have fewer abilities, he or she may have higher skill levels and therefore lower mana costs, shorter cooldowns, or more reliable effects.
So what prevents a player from skipping straight to the top tier? First, there's game enjoyment. A player should be sufficiently interested in the game that they willingly spend some time at each tier, checking out the scenery and learning their new skills. In general, they can move on whenever they wish, though there would be some quests involved. Some of these would be designed to test the players' skills, so one must be familiar with the class and what it can do before moving on. An ideal test would be independent of equipment, so one cannot merely twink past it.
This isn't really so revolutionary. Consider one of the most successful RPG franchises of all time: Zelda. Does Link ever level up? Nope. He gains new abilities periodically, in the form of equipment, generally tied to each dungeon. He does add hearts, which is similar to leveling, but this is not necessary. Some players take pride in finishing a game without picking up any hearts!
Tiers need not be linear. Warriors may be able to enter tiers like Mercenary, Bandit, or Watchman, which lead to Pirate, Gladiator, Sheriff, Captain, or Marksman, and on to Baron, Hero, Paladin, or Defender of the Realm. This takes us close to a WFRP-style Professions system (which I happen to like), especially if they are not arranged in mutually exclusive trees, like class talent trees, but overlap and connect: both warrior-based characters and rogue-based characters could enter the Pirate tier; maybe sorcerers could, too. In this way, a character could develop many sets of skills, though he or she may never be as good as a specialist.
To prevent a player from learning all skills and becoming boringly self-sufficient, we could have the cost of learning a new skills based on both the character's level in that skill, and also his or her total skill points across the board. Thus, while one can reach mid-level in several trees at about the same total cost as mastering one tree, to complete EVERY tree would be ridiculously expensive. We could also have skills decay slowly over time if unused; unless the player followed a bloody-minded system of using all skills in rotation, his or her less-favored skills would not stay maxed out.
Stylistic Development
Reward completionists with more cosmetic or marginal changes. Perhaps there are different schools of magic scattered around the world; one teaches red fireballs, one blue, and one teached lightning bolts. The differences may be nothing but color, or there may be subtle additions which don't threaten game balance, rather like WoW's talent points: red fireballs cause a little DoT, blue fireballs can cause a brief stun effect, lightning bolts do extra damage against armored targets, etc.
Having hidden minor bonuses like these would give sage types a good reason to spend a little time browsing through libraries. Perhaps if a player logs out in a library, there's a chance that his or her character makes a discovery while offline.
A widely-traveled character will have more tricks up his sleeve, and can have a unique exotic look that will turn heads. This isn't limited to magic! A warrior from one region may dual-wield scimitars, while another carries a longsword. There are slight in-game differences - speed, armor piercing, parry bonuses - but they're about equal overall, so the choice is the player's.
Miscellaneous
In most MMORPGs, abilities (strength, stamina, intelligence) rise with level and equipment. More realistically, they should develop from use just like skills. As a bonus, how about ability increases as quest rewards? Complete a foot race to earn a point of agility; solve a puzzle to increase your intelligence. This helps side-questers stay competitive with grinders.
When new content is added, with new ability- or skill-rewarding quests, this could attract a horde of players who want to nuke all the lowbie quests for the ability points. To reduce this, quests could give reduced awards for players that already have high abilities. Winning that footrace in Podunk, a pig farming the village in the remote foothills of Nowhere, should be no great achievement for a master thief. He or she should earn reputation, but not a bazillionth point of Agility just for besting some peasants.
Equipment
Polearms were historically the most popular weapon of all time. Spears or pikes were the main weapon for most historical armies. This is partly because they scale well: a group of spearmen is a tough nut to crack, while a group of swordsmen have to be careful not to hit each other. Still, they should be a valid choice in-game. For example, some kind of staff weapon would be ideal for a warrior who wants to hold a gap, such as a traditional tanking fighter. It could be as simple as giving long weapons a wider zone of threat.
Reducing the Importance of Gear
Most MMORPGs are heavily equipment-oriented. Let's minimize this, scale it way back until having better gear gives you only a modest advantage. This would increase the importance of skill and teamwork, and reduce the importance of shear time spent playing.
Part of the problem is that fantasy worlds seem to be absolutely awash with weapons. You can't walk through the forest without stumbling on a dozen swords, four maces, and an enchanted longbow. Equipment would be much more precious if it was less common.
Decouple equipment from appearance, to some extent. A player should be free to go for a particular style without crippling his or her stats.
This could start with simplifying equipment, especially armor. We never felt there was a need to track left and right gloves separately (with one unusual exception coughDwarf Fortresscough); do we really need to track shirt, belt, gloves, pants, and boots? How about one "armor" one "shield" and maybe one "boots"? Perhaps there would be a handful (no pun intended) of gloves and belts as uncommon items, while most characters use the default "comes with the armor" accessories.
Named Equipment
To help keep equipment interesting, and reduce the need to constantly churn gear, allow a character to name his equipment. Oh, not EVERYTHING, and not often. But if I get a really cool-looking dagger, I should be able to say "this is MINE! I dub it Stabsalot!" OK, yeah, there should be a decent system of randomly-generated names, for the creativity-impaired. Once named, a weapon (possibly armor too) would then grow with the player, making it resistant to obsolescence for some time.
There should be limits, of course. Even a grand master assassin's iron dagger should not be as good as a high-quality mithril dagger, so eventually Stabsalot will need to be retired. But for a span of a tier or two, maybe three, the player has the option to voluntarily step off the equipment treadmill. And then you can put your old weapon over the mantle and let it remind you of your early days.
Backpacks
The backpack is an accepted abstraction, defying physics but probably necessary for a practical game. One small addition would reduce the absurdity somewhat: a "folio," like WoW's key ring, which could store paper objects like shredder operating manual pages, scrolls, and messages. This would make scrolls much more practical.
Economics
The economy in WoW is out of control. Not that it every really was in control, but the Monty Haul syndrome just exploded with the first expansion.
How can one keep an economy balanced? In part, by having active NPCs. First, NPCs should provide some, probably most, resource collection. Let prospectors and miners wander the land, digging up surface deposits and opening mines on their own. Naturally, monster raids will threaten, giving the players a reason to come help out. But aside from that, ore should get mined and swords should get forged without requiring player involvement. This means having NPCs that buy and sell on the auction house, or at minimum from each other. The heroes should focus on hunting monsters, and leave farming for the peasants.
Second, NPCs should own, buy, and sell all kinds of equipment, not just vendor trash. If a player sells his old gear to a shady dealer, he may well run into the pointy end later on when a gang of bandits buy or steal it from said shady dealer.
To prevent hording, NPCs with lots of equipment should be vulnerable to burglary, from PC and NPC rogues. Indeed, PCs own inventory should be valid targets, to some extent. One's equipped gear and most treasured items should be immune to larceny, but not all of a large horde should remain ring-fenced forever.
The Auction House
The auction house can be a lot of fun. It does require quite a bit of micro-management to use effectively, however, and some people just couldn't be bothered. There are at least two things we could do to help.
In addition to "pushing" products, let players "pull." Say Bob the wizard needs an enchanting rod. Rather than haunting the AH for days or hollering into chat, he posts a WTB on the auction house, with a quantity desired and a price. Fred the smith likes the price, forges a rod, and satisfies the request. There could even be a timeout period in which another player - or even an NPC - could underbid and provide a rod for less. If we wanted to be nice, we'd let the crafters "pledge" at the price given, and only actually consume resources if they win the bid. Pledging would make the materials unavailable until the bid was resolved, so only serious offers would be entertained.
Crafting
Personal Workshops
The great orc capital city of Orgrimmar is an impressive place, bristling with spikes and burly warriors. But it's a pain in the ass to make weapons there. The bank and auction house are clear across town from the smithy, so if you forget to pick up heavy thread from your bank account you have to march for ten minutes over and back before you can resume crafting.
So how about personal workshops? You step into the smithy and go to either your own personal door or a generic "your door, wink wink" which zones you into a little workshop. It has an anvil and storage for all your supplies. You can upgrade it for more space and additional facilities, even apprentice craftsmen.
Armories
Beginning crafters often have to make large quantities of mediocre goods to level up. This is generally sold to a vendor at a loss, or posted to the auction house repeatedly until it sells or the player gets fed up. So how about a community armory, where newbie crafters can donate their manufactures and newbie players can collect free equipment? Making donations could also grant a little rep to the crafter.
Such armories should probably be local, rather than universal. This would help maintain the flavor of each zone, so the middle-eastern flavored city will have an armory full of scimitars, the dwarf fortress is full of warhammers, etc.