Hacker News new | ask | show | jobs
by Animats 4282 days ago
My first thought was "Wow, somebody made that work!" After reading their site, it's clear they have a long, long way to go.

In the early days of There, Inc., I was recruited by them, and was in on some early design discussions. They wanted to build a planet-sized 3D virtual world with physical simulation, with all the scaling problems that implies. My comment was "Now let us all join hands around the world", which produced groans. (Doing physical simulation across a cell boundary is hard enough, but if the users can force you to have to do it across N cells, you have a real problem.)

For that system (and for Second Life) the assumption was that all the servers were in the same data center, with low inter-server lag. Trying to do this with home servers and lag in the 100ms and up range is going to suck. However, a distributed system where all sites are hosted in data centers on the same continent with low-latency connections might work.

Now they have to deal with the security problems, the delay problems, etc. They also need a really big MMORPG to exercise the system. Something like a planet-sized version of Minecraft would be a good test. The rules are simple and users create the content. A tougher test would be to create a simple, but huge, driving game, using Open Street Map, "seamless.gov", and topo data. Then load it up with automated traffic and let people drive in it.

1 comments

Many years ago some friends and I kicked around an MMO on a post-apocalyptic version of the US, full-scale. The actual data involved worked out to less than a terabyte, including topo. Modern hardware has made that part tractable.

The problem is, how the fuck do you make it fun?