It's all about immersion
This part of your question jumped out at me:
I can see two ways of expressing the relationship between players and their PCs: either they act in place of the characters (“I'm Dunwick, I do this…”), or they control them like puppets (“Dunwick does this…”). I've never intended to impose a point of view or another on the players, but it looks like Dungeon World wants me to.
You're right, it absolutely does. Dungeon World has a very definite point of view. "Acting" your characters, immersing yourself in them, diving in as deeply as possible, is part of that point of view.
What does it mean to say that an RPG has a point of view?
All games have two components: a goal and a set of rules or constraints. Or to put it another way, what you want and how you're allowed to try and get it.
Some games have simple, definite goals like "score the most points" or "checkmate the opposing king". RPGs by their nature have looser, more open-ended goals that are partly determined by the players themselves: things like "get a ton of loot" or "unravel the conspiracy" or "have fun role-playing conversations between my character and other characters".
But that doesn't mean that all RPGs support all player goals equally. Even where the rules of an RPG don't make a certain player goal impossible, they can still make it more difficult, boring or annoying to achieve then other goals. (You could try to play a traditional heroic knight-like character in a game of Paranoia, but you'll be fighting the game the whole way.) Some RPGs try to support a wide variety of player goals. Others have narrower goals in mind, and their rules are designed to push players towards those goals. I call this alignment of goals and rules the game's point of view.
So when I say that Dungeon World has a very definite point of view, what I mean is that it has very definite goals in mind, and its rules are designed to strongly push the player towards those goals.
What is Dungeon World's point of view?
Fortunately we don't have to guess what goals the game designers had in mind. The SRD tells us up front:
Why Play Dungeon World?
First, to see the characters do amazing things. [...]
Second, to see them struggle together. [...]
Third, because the world still has so many places to explore.
The GM's Agenda section is also enlightening, and also names some goals that the game designers explicitly rejected:
Agenda
Your agenda makes up the things you aim to do at all times while GMing a game of Dungeon World:
- Portray a fantastic world
- Fill the characters’ lives with adventure
- Play to find out what happens
Everything you say and do at the table (and away from the table, too) exists to accomplish these three goals and no others. Things that aren’t on this list aren’t your goals. You’re not trying to beat the players or test their ability to solve complex traps. You’re not here to give the players a chance to explore your finely crafted setting. You’re not trying to kill the players (though monsters might be). You’re most certainly not here to tell everyone a planned-out story.
If I had to sum all this up in one sentence, I would say that the goal of Dungeon World is to immerse the players in a collaboratively-generated adventure story of vivid characters interacting in a fantastic world.
The rules of Dungeon World are all designed to push the players towards this goal in one way or another. For instance, the "collaboratively-generated" part is encouraged by severely restricting the GM's ability to pre-plan campaigns, giving the GM moves that can generate new plot points on the fly, and giving the players the power to push the GM to flesh out parts of the game world that the players find interesting (through moves like Spout Lore and Discern Realities). All this discourages railroading and encourages shared improvisation.
What does all this have to do with addressing characters instead of players?
The real question is, how does this rule help to push players towards the game's goal?
Let's start by reading this part again:
If you talk to the players you may leave out details that are important to what moves the characters make. Since moves are always based on the actions of the character you need to think about what’s happening in terms of those characters—not the players portraying them.
So it's not just about doing a word search on your sentences and replacing "Tony" with "Dunwick". It's about forcing yourself into the mindset of the characters. What do they know? What can they see and hear around them? How does that drive their actions? If you are actively thinking this way, you might realize that you forgot to mention the lead goblin is wearing extra armor, or that the friendly shopkeeper is a Tremarian (Dunwick is racist against Tremarians)...or on the other hand, you might realize that the archer you were about to mention is sitting motionless in the shadows 100 feet away, where Dunwick would have no chance of spotting her. You the GM become more invested in the story just by thinking this way, and with fewer hiccups and less metagame-y infodumping, the story becomes more believable for everyone else.
But even if it was just a simple word-swap, it would still help the players get immersed in the story. Imagine watching a play where the actors keep saying each others' names instead of the characters' names. Changes the mood, doesn’t it? It emphasizes the artificiality and "play-ishness" of the play, which puts a distance between the audience and the story. Dungeon World doesn't want to create that kind of gap. You and the other players are actors in a sense; you are all performing your part of the story for each other. Using the players' names is a distraction. Using the characters' names, on the other hand, generates a continuous string of little near-subconscious reminders to focus on the story and characters, which is what the game wants.
What happens if I break this rule?
Not much, honestly...assuming you are only breaking the names part specifically, and are otherwise doing what you can to preserve the story-focused frame of mind that the rule encourages. (Thanks to Dave for pointing out this nuance.) Alex P is right, this rule by itself isn't a huge one, and leaving it out won't kill the game...but by the same token, leaving it in won't kill you. Why not give it a second chance? Now that you understand some of the reasoning behind it, it might grow on you.
What if I don't even want to immerse myself in the story?
If you and the game disagree on goals, you're going to have a bad time. I don't think this is your case, but if it is...Dungeon World may not be the right game for you.
A (Practical) Cast of Thousands
On the subject of group size, I would highly recommend not going above four regulars your first time out. Dungeon World is a game that demands a lot of the GM in play - you have to constantly keep in mind everyone's capabilities when you're pitching the setup to them and when they're pitching their actions to you. I've been running it for years now, off and on, and I can maybe take 5 people, or 6 on a very good day. You'll get a feeling for what you're capable of as you go, and a regular group can often entertain each other in ways that successive slots at a convention, say, just can't.
However, if you're anticipating a peripheral group who have some sort of interest in sitting down to play on occasion, I'd consider establishing a Crew, from the sadly-never-released Pirate World. The game was kickstarted and PDFs distributed to backers, but I can't find anywhere to buy it. (If anyone knows otherwise that'll be a nice surprise.) In a game of Pirate World, Crew are more than hirelings but less than PCs, who accompany the PCs regularly on their adventurers. In a game more based around a hub town or towns (and a notable Dungeon or Dungeons), Crew can be adapted to be the colorful townsfolk and friendly monsters and animal companions and such that you may be familiar with. They're statted out about like this:
Innismore, Irascible Alchemist
Hunger: Highly Illegal Ingredients
Benefits: Know something vital about a powerful adversary
Mix up a potion to overcome some obstacle
Loyalty: 2 Resilience: 1
The Benefits are the things they can do for you. When you want one of your Crew's Benefits, spend 1-Loyalty to get it, but first roll +Loyalty. On a 10+, that's the end of it. On a 7-9, the GM chooses 1; on a 6-, they choose 2.
- You must satisfy their Hunger before they'll help you again
- You must Defy Danger or otherwise put yourself at risk in order to make use of the Benefit
- It costs 2-Loyalty instead of 1.
Benefits are very conceptually similar to monster moves; that is to say, they don't have to be anything more than descriptions of the cool thing that you want the creature to do. When the PCs take advantage of them you'll usually have some kind of context to flesh them out (in this case, the obstacle or adversary PCs need to consult Innismore about). You can also define Benefits more strictly, as something to guarantee the PCs they wouldn't otherwise be able to find:
Whitecrush, Lizardfolk Shaman
Hunger: Forgotten knowledge of magic
Benefits: Provide access to a lizardfolk safehouse (a safe place to Make Camp for one night, and folk medicine to cure one debility)
Dispatch a young bravo as a guardian (treat as a hireling with Cost: glory in victory; split 7 points between Warrior, Protector, and Loyalty)
Loyalty: 1 Resilience: 2
The Hunger is the troublesome thing they need PCs to provide them with. When you satisfy your Crew's Hunger out of more than a sense of obligation, they gain 1-Loyalty. Hunger is something to keep in mind when you're offering PCs opportunities - sure you could destroy all the cultists' horrible reagents, but maybe you can smuggle them in to Innismore. Sure, Wizzrobe could copy their ritual notes into his own spellbook, but they'd also be worth bringing to Whitecrush. That sort of thing.
Loyalty is most easily altered through asking for Benefits and feeding Hunger, but it's there to represent how this person feels about the PCs. Doing things they would appreciate or putting them in danger can also bump it up or down, but it's proper sportsmanship to tell the PCs the requirements or consequences if they're making a decision and loyalty is on the line.
Resilience is a rough measure of how much damage they can take, equal to about 1 for every 5 full hit points they'd have as a monster. Each significant attack takes off 1 resilience, and if they get whacked at 0 resilience left their fate is completely in your hands.
Improving Crew is certainly possible over time, as PCs establish more of a relationship with them. If it seems like they'd gain something new, feel free to write some of your own moves that key off Loyalty to reflect these new capabilities - Whitecrush's bravos might get more points to distribute or she might develop some rituals of her own, or Innismore can kit the PCs out to "field test" some alchemist grenades (roll Loyalty and you get hold on low rolls that you can spend to activate complications).
When someone on the periphery wants to jump in for a session, they can pick up one of the Crew. They can use the main benefits as normal - they make the choices about costs instead of the GM - and they can also roll +Loyalty to help the main PCs in other ways, anything an Aid roll might work for. On a 10+ the PC they're helping gets a +2, on a 7-9 a +1, and on a 6- in addition to what the GM says, they feel obligated for letting everyone down and gain 1-Loyalty. In general they won't be exposing themselves directly to danger, but you've got Resilience for those cases where they get caught out.
Best Answer
I can't imagine running any campaign that isn't improvisational to some degree. I am running a Dungeon World campaign which is more traditional in campaign design because my players come from more of a D&D background, so I'm working from that perspective. That works fine- effectively my fronts are running in a sandbox and the players get to choose where they go and what they do but they often look to me to give them a direction when they aren't sure where to go next or what to do. My goal is not to restrict the agency of the players or the power of the moves, however - if anything I'm trying to guide them into more of a DW mindset - but the consequence of the set-up is that the game we are playing is carried more on my shoulders as a GM in narrative and worldbuilding terms rather than the constant collaborative efforts that characterise the game in its purest form.
I think the biggest element to this is the fundamental guideline of "Draw maps, leave blanks" - I apply that to everything I am doing as the game develops. I know in general what might happen, I know what the various fronts and groups in the world are doing in general and then we learn the details as the PCs interact with them. Then I can set up problems for the players to solve without knowing how they are going to actually solve them- that is where the improvisation really kicks in.
To look at the things you are interested in:
So I think you can find the things you are looking for in this family of game, but if you want more control then maybe you should look for a game that offers you that by design- there are plenty of them around.
(As an aside, we're podcasting this game, so you can hear how it goes from the link on my profile, but bear in mind that we're basically all idiots, so although hilarity ensues it's probably the opposite of a textbook guide to running or playing Dungeon World. )