Over the past few weeks, there’s been a little more chatter than usual in the mainstream media and on blogs about 3D virtual worlds and their relative merits. Both Google Earth (GE) and Second Life (SL) have been mooted as a paradigm for the budding metaverse.
Currently, SL is much more of a full-service virtual 3D space, with avatars, one shared “state” of the world, scriptable objects and an economy. Google lacks all this, and if it were to want to become a competitor in the metaverse market, it would need to add that functionality.
A more fundamental, persistent difference between GE and SL, however, is that GE aims to be a mirror of the real world, whereas SL is a made-up world, with a geography dictated by the whimsy of its users.
[In the rest of the post, I argue that there are two conventions we can let go off when building virtual worlds: That space has to be flat (Euclidian) and that the topology has to follow the geometry of space (it can be non-contiguous, like the web.) It just takes me a while to say that:-)]
For GE’s developers, then, the main aim becomes accuracy in representation. For SL’s developers, the challenge becomes giving users free reign to express their creativity in shared 3D.
The reason GE is a globe is because Earth is a globe. Paris is to the southeast of London in GE because that’s where it is in real life. SL has no such constraints on the organization of its content; and yet its creators chose a very conventional world: the infinite Euclidian plane, with disk topology to match — a flat Earth. On it, we are exhorted to buy property — first come, first served, with the desirable centralized locations in scarce supply, and bound to increase in value as demand increases. It’s remarkably like local real-estate markets on “real” Earth.
But why should SL’s geometry and topology be anything like what’s found on Earth, when there are so many other options available?
I think (in the nicest possible way) that there is something of a failure of the imagination to SL’s approach. It reminds me of how for the first three years of the world wide web, we produced content as if we were still making newspapers and magazines, with virtual “issues”. Slowly, it dawned on us that the web is not made of paper, and we began adopting dynamic, just-in-time models of publishing, native to the web, of which the blog is its most evolved form.
It took time for us web users to abandon the old metaphors that worked best in old media. We should now leave behind conventional spatial notions when building the metaverse. We need to give the metaverse a native form, free of real-world constraints.
Before imagining what such a place might look like, a quick aside about World of Warcraft. It too is set on a Euclidian grid, mimicking a medieval-type map. Yet traversing the geography of WoW is fun. In SL, after an initial bout of newbie exploration, it quickly becomes boring (and time consuming) to travel conventionally; people teleport instead. What’s the difference between WoW and SL? WoW is a game, a goal-driven environment, whereas SL is not. In WoW, Euclidian grid topology is a feature of the game play. In SL, it is an impediment — a scarcity engine.
The widespread use in SL of teleporting to travel should tell us something — that the strongest connection between spaces in a virtual world is not geographical (as it is arbitrary) but semantic. The world wide web was a revolution for precisely this reason — it collapsed physical distance, letting you “teleport” between related sites with each click on a link. That kind of topology proved to be the most natural for navigating all information where the geospatial component is not its most important attribute.
Metaverse 2.0 should incorporate the best of both worlds: It should combine the efficiency of a node-like topology on the macro level with 3D realism at the micro level. You want your closest friends to be, well, closest. You should be able to walk along the nodes that constitute the six degrees of separation between you and everyone else.
How? Imagine a 3D MySpace “homespace” (thanks Avi) where you and every other member are given a hexagon onto which you can build your own “home”. You then find six reciprocating “best friends” as neighbors. These neighbors’ hexagons become the immediate surroundings of your own hexagon, and indeed they are all visible in the distance from your own virtual mansion.
Yet these best friends of yours are under no obligation to choose compatible best friends as neighbors, other than yourself. If you share one or more “best friends”, then you can certainly create contiguous neighborhoods between your hexagons, but that’s not a requirement.
Your six friends’s properties will never be more than a short walk away. If you walk to one, you may notice that its adjacent properties don’t match yours. But your neighbors’ neighbors may in turn be your neighbors again, at a different angle. The upshot is that you can navigate such a world via loops and clusters of nodes, rarther than a predefined grid topology.
Why hexagons? It’s a shape that can tile a Euclidian plane. There are other tiling solutions (a pair of Penrose tiles, for example), though repeating hexagons have the most neighbors.
But why should Metaverse 2.0’s geometry be Euclidian? A sphere’s geometry is elliptic, though space is “crimped” as you travel. There is another possibility, one that offers “more space” and hence the potential for more neighbors: The hyperbolic plane. If a virtual world subscribes to hyperbolic geometry, then the tiling pattern can incorporate both hexagons and heptagons, like so:
(If this is getting too abstract, please read Crocheting the Hyperbolic Plane, a wonderful piece of expository writing from which I took the above picture.)
In a hyperbolic virtual world, the heptagons could become the commonses around which likeminded people cluster. The hexagon-heptagon tiling pattern is just one of many possible combinations, however (top row, third from the left, to be precise).
The future metaverse doesn’t need to look like any of these example, and likely won’t. (Personally, I want to reside on the surface of a Klein bottle or Möbius strip.) I do think we should abandon the arbitrary constraints on topology and geometry that current virtual worlds labor under.
Further reading: While I was writing this, Avi Bar-Ze’ev asked: “Can we come up with a world that’s as rich as Earth but unconstrained and content-driven, like the topology of the web?”. Mark Wallace posits a “distributed metaverse”, where, presumably, we all power our own 3D spaces, but connect them to each other via agreed-to protocols. I think that’s the way forward.
[Update 2006-06-04 08:49 UTC: Avi has already continued the debate, here.]
[Update 2006-06-05 07:21 UTC: Cory “Linden” considered topology and geometry when developing Second Life. More here.]
How many beers did you have before you wrote this? :-)
Seriously, that’s some pretty deep thinking. Cool ideas with the geometries. But, I’m not sure it would have mass appeal. I don’t think GE will be metaverse 2.0, its the realverse.
But, you could certainly use the same application to provide access to a different database which would be metaverse 2.0 if you wanted. Just come up with an alternate Earth, generate your own simulated aerial photography, DEM terrain, 3D buildings, etc. You could still use network links, point to the web, etc. You would need a separate coordinate system, and KMLs would have to know the difference between realverse and metaverse versions of Google Earth. You could also use SketchUp to create 3D models, and the Google 3D Metaverse Warehouse to populate it…
I’m not sure there’s a good reason for different users to have a consistant geometry either. Your hexagon could be my 6 story building, could be Frank’s teleport pad.
The underlying data structure of the linked metaverse should be seperate from the display metaphor that my metaverse browser uses.
> I think (in the nicest possible way) that there is something of a failure of the imagination to SL’s approach.
> But why should SL’s geometry and topology be anything like what’s found on Earth, when there are so many other options available?
There are a few interesting things in the development of the idea of Second Life and the kind of spatial metaphor there. They may be helpful in thinking about this evolution.
As the story goes, when SL started Linden Lab was making hardware, like VR goggles to jack into virtual space, a very first generation idea of the metaverse. They started making a world to look at (then called Linden World) and realized that letting people build the world was more interesting (and saves on all those ballooning art costs that plague the virtual world industry elsewhere :). Then later they moved onto the idea of residents owning what they build, so there’s been an evolution of the SL vision from VR interface to social world to creative/economic engine.
Similarly SL started out as totally contiguous and shared, and this was seen as essential to its development (and may well have been and still be important–there’s a social engineering component to all of this to help it take flight). There was only one main body of land and for a long while you could not teleport directly anywhere you wanted (you had to navigate a series of “telehubs” arriving at the one nearest to where you wanted to go–this was to create density of people and content and promote value in real estate location). Later you could purchase your own private island, and we’ve seen a flight to these. I can’t find the link to the pictures offhand, but the evolution from mainland to islands from 2003 to the present is worth 10,000 words. The next step looks to be the hyper/networked space where geography won’t matter at all.
But since 3D spaces are individually geographical, I wonder as you do what kind of combinations of Euclidian space and networked space we’ll see. One idea might be a dynamic mainland composed of all the most popular places (generated by time people spend there and tagging systems), so you can get the topsight view of the most interesting places all in one glance, helping you navigate what you’re most likely to like easier than just links. Another idea might be to build your own map. Say you like 30 different virtual spaces, you could drag and drop them onto a map of your liking, arranging the geography to your tastes. If I walk off the left side of my sim I’ll go to my friend X’s sim, but if someone else had my sim on their map, they might walk off the left and go somewhere else. Not sure, but a couple of ideas there.
I’ve been banging the metaverse and Second Life meets Google Earth drum for a while and have the opportunity to talk about it at Where 2.0 coming up real soon. The talk is called “Second Life and Google Earth Mashups: Virtual Worlds Meet Geospace.” Great discussion here. It’s all going in the talk! :) Now I’m perfectly happy to be your avatar for that, ha, so speak through me. What else do you think I should be saying there?
Of course, there’s an easy non-topological explanation for that as well – the SL mainland is full of crap. In areas that aren’t full of spinning ads and depressing block architecture, it’s actually quite fun to travel. In fact I have talked to several people just today who find that just riding around on public transport in the right sort of area is enjoyable in itself.
Ordinal > Of course, there’s an easy non-topological explanation for that as well – the SL mainland is full of crap. In areas that aren’t full of spinning ads and depressing block architecture, it’s actually quite fun to travel.
Totally. What do you think of the idea I mention above?
“One idea might be a dynamic mainland composed of all the most popular places (generated by time people spend there and tagging systems), so you can get the topsight view of the most interesting places all in one glance, helping you navigate what you’re most likely to like easier than just links.”
In other words you could travel across a mainland made of the best stuff, however that gets defined.
A world where macro-topology is determined by tagging or whatever system is used is interesting, certainly… it’s just one that I’m not sure I’d like to see in practice.
The thing about geography is that I know it’s independent of the opinions of people who inhabit it. I can fly across a landscape and, while there might be certain restrictions e.g. waterfront land is more expensive so will often have certain sorts of owners who are prepared to buy it, it gives me an alternative to whatever popularity metric is being proposed.
Given that every time a new metric arrives there’s an immediate struggle to work out the best way to game it, and that popularity systems in SL exist, I would rather have the option of also having a… well, not entirely random but less gameable way of seeing new things.
There’s also that, to be honest, if a 3D world is going to be immersive and work on a predictable Euclidean type model on the small scale (within sims), I’d like it to work on that basis on a larger scale as well. Otherwise I lose all of the immersion I’ve built up before moving. Within a sim I can be sure that flying from one point to another will cross consistent space – moving between them, I would like to have the same.
Particularly as I build automatic vehicles that would really be confused by a dynamic system. How would that cope with bots?
I like what I read here. I’d love to choose a time in the past, say 1600s, yet visit a real place, say Williamsburg, VA, and meet ‘created characters’ from that time period. Maybe we could ‘re-create’ our history and play people from the past — travelling through both time and space. We might choose to be a surveyor who opens up the west, and encounters other ‘re-created’ characters — using a google earth based on both space and time. Some of us might indeed decide to visit the future as well, with the impacts of global warming, pandemics, and new discoveries.
Stefan, take a look at Cory “Linden’s” explanation of how the SL grid evolved. It should be in your trackback, or check the bottom of my blog post on the subject for a link.
WOW is better to fly around in because it’s not streamed to your computer, it’s stored there already, which makes for smoother transitions. BUT, does WOW offer space that is not a landscape in the normal sense (dungeons don’t really count)? I kinda don’t think so — seems pretty euclidian to me.
Another point — although all the land in SL seems to be flat, promoting a flat-earth feel, when one looks at the map, even though it is similar to other maps you might have seen, there is nothing that says those spaces are really contiguous, aside from the representation of it. The lattice/grid of points is more applicable as a spatial theory for virtual space, but much harder to immediately make sense of for the average user.
And of course, as Ordinal alluded to, areas with dissimilar physics or that are dynamically different from each other will not engender lots of movement between them.
It certainly sounds interesting, but I think that humans, being human, are sort of used to Euclidean geometry already, and breaking it up would be more trouble than it’s worth. I can’t even find my car in the parking lot sometimes.
A potential compromise would be to situate those hexagons as floating islands within a Euclidean space of infinite dimensions, confined to a hexagon grid and freely teleportible, such that one would require the permission of another island’s owner to teleport adjacent to their island (and by extension join to them), or within a certain vertical distance above them (to avoid the possibility of e.g. using large clumps of islands to blot out a particular user’s sky).
I just popped back here to say that I got interested in the idea of sim tagging from Jerry’s comment, and so built a little HUD device that lets people tag sims within SL and puts together “people who liked sim X also liked sim Y” data.
Excellent article and a fresh perspective. It may be interesting to see how this evolves in ten years. I think, even if we totally let go
of all this planning, web3d will somehow evolve
and mutate, ANYWAYS. THIS evolution I would like to track :)
This is just a little to deep and heady for me, but what little I could understand was fascinating.
If Euclidian grid topology is an impediment to accessing information, could this be the reason why splash pages on websites are super annoying?
Splash pages are a way for “flat world” designers to impose flat world concepts on the web. Along the lines of “Before my customers enter my shop, I want them to see this message first.”
Splash pages are an attempt to enforce a prescribed journey in a world (the web) that does not subscribe to prescribed paths.