So, what is it?

• It's role-playing.
  Several people get together in a game of make-belief, portraying a character of their choice -- a leading role in the unfolding play --, creating a story together. It's a bit like radio drama, only they're making the plot up as they go. (It would be Improv^°isational Theatre or the Commedia dell' Arte if people actually stood up and did things rather than just talk about them.)
  
  
• It's dice-less.
  Unlike in many other games, no dice are used. Coincidence and surprise are either introduced by more complex randomizers, like Tarot cards, or, more often, by the sheer unpredictability of your fellow players.
  
  
• It's rule-less.
  There is no explicit or elaborate rule-book used to decide what happens. What can happen is limited only by the collective imagination of the players, and by agreement of what is acceptable, often based on the conventions of the genre portrayed (May player characters kill each other? How is violence portrayed? etc.).
  
  
• It's storyteller-less.
  This does not mean that DRASTIC RP always is without narration, it just means that any player can narrate or introduce supporting roles any time she chooses (as long as she does not openly contradict any fact already established by another player).
  "Storyteller" is a word borrowed from White Wolf; in our case, everybody is helping tell the story (so the distinction becomes meaningless), and there are not necessarily "referees" or "game masters" who "own" the story or the game-world.
• It's in character.
  Some people would call it a dramatist's game. The focus is on enjoying the playing and exploring of the role one has created, and/or the unfolding of the story one is telling with others, not some kind of "winning the game." The more one gets entangled in the story, thinks and feels and experiences the game-world like the ones character would, the more in character (IC) one gets. People who for the duration of the play almost forget that they are "but actors" are called "deep IC."
  
  

In Everway, each actor chooses a couple of images that she can then combine into an ad hoc background story for her character ("This picture shows my aunt's friend's sister as the grandmothers of invention sneak up to try and steal the secret samadhi from her. This next one shows my character Moon in her Eyes as she travels through the Swamp of Despair on her way to said aunt..."). While your gameshop might still have Everway "boosters" -- sets of cards with pretty fantasy pictures by artists of renown, with inspiring questions about the depicted scene on the back -- which I recommend to those playing fantasy, any pictures that you find inspiring will suffice. (They are not like Pokemon or Money: The Gathering boosters; the only in-game advantage they offer is that of a more colourful background.)

I encourage players to experiment with this way of defining a role if they aren't already sure what they'd like to play. If they have already come up with a character (or adapted one from movie and book characters), that's perfectly adequate.

Some people like to jot down some notes about their characters. In Hero Wars, a deductive way of character creation is suggested -- the actor writes up a background (perchance one inspired as described above), then she -- maybe with the other actors and actresses, maybe with the director/game-master, if there is one -- underlines the keywords of this story. These become the descriptors of her role (or character).

Over the Edge requests only four descriptors for a role: The primary descriptor is what defines the character; it is often the character's occupation. The second and third might be a hobby, inclination, or generally anything that the character spends a lot of time or thoughts on. The fourth is some quirk or flaw to make the role more interesting.
I encourage players to enrich their roles with as many of those "lesser descriptors" as they see fit.

If there is no storyteller, the basic principle is first come, first serve -- whichever actor defines a detail of the game-world (or setting, or stage) first is right by virtue of beating the others to it. The others may elaborate on this, but they usually will avoid openly undoing someone else's creation. Exceptions from this rule ("Upon closer inspection, [other actor's character] notices to her horror that the X is not an X at all, but a U!") should be used sparingly and with great care.
This way, an actress can decide whether her character succeeds in an attempted action, like picking a lock, or she may leave the decision to another member of the cast. Surprisingly, this does not take the thrill out of the game for most people -- succeeding (managing to open the lock) may turn out to be something bad, after all (perchance releasing unspeakable horrors, which would be bad for the characters, perchance just leading to a less satisfying story, which would be worse for the actors). Often not playing some infallible superman just renders not only more intriguing characters, but also a more satisfying story -- your mileage may vary.

Those reluctant to implement all DRASTIC principles at once may be interested in trying a "less drastic" form first:
There is a director (or narrator) who "owns" the story and the setting, but actors may temporarily take control over the story from the narrator. Theatrix offers two ways of doing this:

1. Each player gets a number of plot points, and/or the cast (all players together) get a pool of plot points. When a player spends one of those points, what they state automatically becomes a fact in the story ("Hey Pete, I examined the trail in the snow: the murderer's got a limp!").
   This technique works better with situation-based than with plot-based narration (see below) as an actress may inadvertantly spend a plot point on contradicting some of the plot or facts the narrator had in mind. The less flexible this plot is, the more difficult it will be to adapt it to accomodate the changes introduced by the actors' plot points.
   
   
2. A player burns a descriptor. If a character has the descriptor librarian (Giles, anyone?), the player can reasonably argue that the character will automatically succeed when trying to do something library-related -- Marcus Brody excepted .
   If the player however needs something akin of a deus ex machina ("What a stroke of luck, I took exactly the book we need on this trip to the desert!"), they need to cross the descriptor off their character notes so it can not be used in such a fashion again. Use with taste.

When a decision isn't preempted by burning a descriptor or by using a plot point -- either because the actor chooses not to, or because you are not using those principles at all --, you will sometimes have to decide what's possible and what isn't.
I usually reckon that if an actor can provide three good reasons why something would work ("I throw a fireball at the prince. I suppose I have a pretty good chance of actually hitting him with it, as I can use the fiery essence of the fireplace, have studied fire magic for three long years, and as the prince is a pretty much out in the open where there's no cover."), he has a pretty strong case going for him -- unless someone provides three equally compelling reasons why his action cannot take place as he imagined. (This case tends to be rare in my casts. If it isn't in yours, the Matrix system^° for formal arguing may be for you.)
If there is no dispute over the idea, the narrator will decide what happens as a result of the attempted action, taking into account such factors as the mood/genre of the story, how probable success would appear, and what would be desirable for the pace and dramaturgy of the story.
Finally, the narrator may choose to use a randomizer on top of those considerations. I have found it increasingly useful to not use dice, which tell you whether an action succeeds (and ideally, how well the character did (or looked while doing it)), but things like Tarot cards, which give you a hint of how this success (or failure) manifests:

In one session, the characters slept in something akin to an African hut, together with other guests of the wedding they'd attended. One of the characters decided to snuggle up to another person -- one of the supporting roles (who was there with her lover). I drew three cards from the Everway deck, which is specifically tailored to role-playing needs, and ended up with Deceit, Innocence, and Separation, which I interpreted as the person snuggling up to the warmth of our character ("deceiving her lover") in her sleep ("innocence"), with said lover awakening to that scene (threat of "separation") and throwing a major fit.

Of course, the very same scene could have happened after a bad roll, but the cards often "suggest" turns that I might otherwise not have thought of, thereby enriching the story and avoiding repetitive elements.

The more input the actors are allowed to make, the less certaintity there is that the story will at all resemble what the narrator (or narrators) had thought of. Where a traditional storyteller may create a plot, having an idea for a "complete story" with a beginning, some middle bits and an end, a narrator that affords her actors and actresses the freedom to drastically alter the script "while the cameras are running" is unlikely to see the story unfold or end like she had envisioned.

To avoid having to "push the actors back on the right track" of the pre-destined storyline (derisively called "railroading" by some), the paradigm of the sitcom (where situation and characters are "everything") seems more suitable to DRASTIC RP than the fable (where the characters are stereotypical to a fault, and plot and outcome are everything), but of course, it's not all black and white. The director can maintain some sense of control if she prepares some building blocks (situations, locations, strong and memorable supporting roles and antagonists) that she may use, alter, discard or rearrange according to the players' actions. These add to the uniqueness of the story and can be prepared in advance, while still affording the director maximum flexibility later on. Thus prepared, the director not only has nothing to fear from capricious actors changing the story, she can actually enjoy the changes as now, the actors are not simply playing out what she had already envisioned. She instead becomes a part of an all-new story.
This works for one or several narrators, as well as for actors.

Toon offers "random situation charts" for many genres that let you create a simple situation (or plot premise) with a few rolls of the dice. (Amazingly, this seems a good enough mechanism to reproduce all Star Trek plot premises known to man.) Throw in a few interesting characters and supporting roles, and feel the story unfold!

Theatrix offers some tips on how to handle action resolution to create a certain pacing (and, if desired, turn of events, providing subtle railroading) for said situations which come in handy if one desires to create a quasi-cineastic dramaturgy, a "moviesque feel."