Last month, I had the pleasure to co-organize a small event in Seoul about digital and physical space, and how technologies reshape them. One of the speaker, Jake Song gave an interesting talk about the evolution of "virtual space" in multi-player games. A South Korean programmer, regarded as one of the greatest game developers in Korea, Jake is one the of the creator of Lineage and is now CEO of XL Games.
He started by describing text MUDs (1978), in which 100 to 200 concurrent players wandered around virtual places in the form of interconnected rooms. He described to what extent today’s MMORPG inherits most of its design (chat, emote, social structure, etc) but more interestingly pointed out how MUD space was "not correct in 2D sense" using the following schema:

The next step corresponded to 2D MMORPG such as Lineage or Ultima Online, which involved 3000~5000 concurrent players per world. Due to the technical impossibility to have everyone in the same place, there are "parallel universes. As opposed to MUDs, space in geographically correct and as he showed with exampled, players approximately needed about 2 hours from end to end by walking. Through various examples, he showed how buildings are smaller than real world and the necessity to have fast transportation methods (horses, teleportation, etc.)

Then, with 3D MMORPG like EverQuest, Lineage II or WoW, game space is designed for 2000~3000 concurrent players. Compared to 2D MMORPG, the number of concurrent players are reduced somewhat because server has to handle more complex 3D data. The game space generally corresponds to 20 × 20 km in size and transportation methods are needed more than ever (mounts, flying mounts, teleportation). Level design included place utility buildings for players to access conveniently. Shops, banks, etc can be placed far from each other to make players feel the city big, but it will make players inconvenient (" Down time means idle time such like staying in town, moving to other place, etc."). The virtual environment is large enough to feel like the real world and is similar enough to use common sense to navigate.
He concluded with the challenges: the difficulty to have larger game world, the possibility to have user-generated content (to populate worlds), the difficulty to have "one big world" and the ever-growing inclusion of environmental change (weather changing accordingly with weather feeds), evolution over time (deformable terrain
destructible building, changing forest, buildings turned into ruins, etc.
So, down the road, the main issues are:
- geographical correctness: should the system looks and behave like a real-life equivalent (which somehwat connects to the work of Harry Drew).
- given that geographical correctness is now common, time and transportation is an issue: it takes time to go from A to B and transportation systems must be designed (teleportation, flying in Second Life).
- presence of concurrent players.
- presence of "places" with functional capabilities (communication, trading).
Although this may look obvious to many reader, this description if interesting from the research point of view (as well as to have the developers' opinion). In my case, this is important for my research about how location-awareness interfaces can convey information about people's whereabouts in digital spaces. Given the differences to represent space, there are some implications in the way location-awareness tools can be designed. More about this topic later.
Thanks for the post. I'd be interested to know if he had design tips to convey a sense of past and present inhabitation, and for affording 'gathering' of people.
NB: Down time sounds a bit like Heidegger's term 'marginal practices'.
Posted by: ErikC | Oct 02, 2007 at 03:19
I wrote something about the different arrangements of space in virtual worlds for the new book Space Time Play. I over-ran my word allowance, so put the full version of the paper here.
Richard
Posted by: Richard Bartle | Oct 02, 2007 at 04:19
This is really interesting stuff. I was thinking about doing a post a while back wondering why most MMORPGs utilize a Euclidean/Newtonian concept of space instead of doing something more interested (like the MUD diagram above). But after thinking about it a bit, I realized that 1) space really isn't Euclidean in fact and 2) it's kind of strange to even describe *simulated* space as Euclidean. Much more important for the game is what players can do with the space -- i.e. space is understood, in part, as an obstacle to be overcome. Teleportation, bindpoints, maps, visual markers, mounts, etc. are all just affordances that player use to deal with the problem of space. However, Euclidean space can be important insofar as it allows players to develop cognitive maps that makes sense of their environment.
Posted by: greglas | Oct 02, 2007 at 09:50
Well put, Greg. I think that virtual worlds tend toward a Newtonian space only to the extent that such space offers the most familiar kind of arena for social action. That is, as more affordances are made available in a space, it tends to approximate a pragmatic version of the offline geometry we inhabit and in which we are used to acting with bodies and with language (or text; i.e., representation). Since most graphical virtual worlds are looking to appeal to a wider and wider swath of users, that trend only intensifies for them.
This makes the Escher-esque space of MUDs sensible, in a way -- the text-based nature of the environment didn't constrain the making and mapping of the space beyond the point necessary for use. It seems to be an interesting combination of constraints in one area (expression through text) leading to open ended-ness in another (space). This is why I become a bit concerned when space is taken as the organizing metaphor for how to understand virtual worlds -- it seems to me more that the proliferation of (Euclidean) space in them is an after effect, and not a necessary one, of how they are domains for performative action.
Posted by: Thomas Malaby | Oct 02, 2007 at 10:08
Thomas, I think that's right -- the MUD diagram is more sensible for the MUD interface.
In a way, MUD space should be more flexible because our linguistic/textual conceptions of space are more easily rendered as performative and nodal. E.g., if we are instructing a taxi driver on how to get to a destination, we give landmarks and routes, just as we instruct the computer in a text-based space.
However, in contemporary MMORPGs, visual rendition acts as a constraint on this mode of mapping space because when space is seen, we want all possible coordinates along the three axes to be filled in with some data (even if that data is non-functional). Any gap would be disconcerting because it would differ so radically from ordinary experience.
Richard -- I just skimmed your paper at this point, but I will get to it soon.
Posted by: greglas | Oct 02, 2007 at 10:25
@Richard: I was intrigued by the vision at the end of your paper. Of a world with many histories which the player could follow. Pretty daunting technically, but it sounds like a great arena for the kind of story based play I enjoy. If players could also have some control of who shared their particular history, then it perhaps each player could be the hero of their timeline, and a bit player in other peoples timelines.
Such a world would have to be designed from the ground up for alternative histories I think. Maybe each geographical space would need a bunch of alternate states. Probably easier to do in text. Did anybody do anything like that in MUD days? I can only think of single player examples at the moment.
Posted by: Hellinar | Oct 02, 2007 at 15:19
Thomas Malaby>I think that virtual worlds tend toward a Newtonian space only to the extent that such space offers the most familiar kind of arena for social action.
It's not a social action thing, it's an action thing. We have the same ways of laying out worlds in single-player games, not just in virtual worlds, and where's the social action in those?
The reason we do this is that if you want to mess with the way that humans internally model the world they're in, you need a very good reason to do so. People are used to their sense of location in the world being updated automatically for them using cognitive faculties delivered through millennia of evolution; if you disrupt that, they'll notice and they'll expect an explanation.
There are a number of differences between the Euclidian space of a graphical world (imposed on it by the graphics) and the conceptual space of a textual world. The first thing to note is that you can do Euclidian in conceptual but not vice-versa; the second is that if you pretty well exclusively use Euclidian, you can more easily exploit some properties of the domain to your advantage (eg. distance vision).
Personally, I prefer a non-Euclidian framework. The main reason is that it allows places to be defined by their significance, rather than their physical size. A mountain can be less important than a corridor; a desert 12-rooms side to side can feel wide without taking forever to cross.
In their heads, people don't model the world in Euclidian space. Think of the layout of your desk. Think of the vastness of the ocean. They both have their cognitive boundaries and fit neatly in your mind's eye, but they're at greatly differing scale. In a Euclidian world, you have to give the players the picture from which to build up their internal model; in a non-Euclidian world, you can give them something far closer to what that internal model will be, and you have a much stronger ability to shape it.
There are other things you can do with a conceptual map too, such as have enormous complexes of rooms in close (in terms of nodes between them) proximity, whereas in a Euclidian space they'd overlap - this is one reason why we have instances in today's graphical worlds. However, you don't get ready-made pictures with non-Euclidian worlds, so in practice none of these advantages are worth anything.
Richard
Posted by: Richard Bartle | Oct 03, 2007 at 03:56
We agree, Richard. The phrase "social action" for us social theory wonks doesn't actually entail the immediate presence of others. It signals what you described -- that the way we (make sense of and) act in the world is at least in large part socally constructed.
Posted by: Thomas Malaby | Oct 03, 2007 at 08:20
In their heads, people don't model the world in Euclidian space.
In navigation you can distinguish between: survey-knowledge (near euclidean, distorted bird-view map) and route-knowledge (go to landmark, turn left, go 100 m, turn right etc). So I think they do both.
Posted by: Ola Fosheim Grøstad | Oct 03, 2007 at 09:56
They do that and more, Ola. How people make sense of and operate in their environment is not only cognitive in the second-order sense, although those aspects of it can be modeled (always in an approximate way) as you describe. As Richard points out, meaning and emotion are part of how we order our world and, furthermore, our embodied practice of moving about itself helps orient us in ways that often do not even need to rise to the level of reflective consciousness (those moments when we pause and more reflectively consult what we know of somewhere). When I returned in 2002 to Chania, where I had done my PhD research 8 years before, I made my way around as much because of where my feet took me, and the affective responses to meaningful places, as through appeals to my sense of the whole and my sense of specific routes.
Posted by: Thomas Malaby | Oct 03, 2007 at 11:23