As for Second Life, they are opening up: the protocols are documented, LL has open sourced their viewer, and there are third party implementations of both the viewer and the server.

There have been a couple of challengers that had better graphics, but they were far more proprietary and they failed.

For an application like Second Life, where things ‘rez’ in-front of the player it would just take too long. Quake Live, for example, does function in the browser but all the maps and models are downloaded ahead of time. So in this way it is not an example of browsing the web and discovering new 3D content. User-created content is out of the question.

However computing power is getting good enough to have almost nothing pre-calculated, which is great. You might be able to get away with just downloading textures and geometry on the fly, especially in the future with faster broadband. However, with advances in computing power, even some fairly convincing textures may be able to be procedurally generated at runtime faster than it would take to download them.

But I would say it’s not there yet for general PCs. You still need fairly powerful graphics cards for dynamic lighting engines. I am certain something like this will come about in the future though.

Why does the world need to be rebuilt?

Yes, Quake will have to be _adapted_ but as is would make a crude start, but a start non-the-less!

The concept is much older than that; it goes back at least to the 1960’s. Snowcrash may have introduced it to a new generation but that’s all.

As for using the Quake engine, I’m not sure it’s all that well suited for virtual environments. Engines like the Quake engine are written for very different kinds of environments and usage scenarios. Among other things, Quake worlds don’t usually have hundreds of remote users, and Quake worlds don’t get rebuilt while levels are running.