An Idea for the Future of MMOs

How many of you here are familiar with IRC, or Internet Relay Chat? That protocol allows multiple servers to link together in an IRC network, which in turn allows users on one server to access channels and users on another server in the same network. This allows IRC operators to split the burden of managing the network across a large area and allow anybody in the world to join in from a nearby server. The only question is... what if we applied this to video games?

There are a few good reasons for this kind of option to exist. For a small company, it eliminates the substantial costs of buying multiple servers (keeping in mind that a good server can cost anywhere from $300 to thousands) to host the game on and shortens the bill down significantly. It also allows players to join in with players on other servers. Similarly to how a client connected to, say, orwell.freenode.net can chat with a client connected to moon.freenode.net, you could have someone on curse.<mmo-name-here>.wtf go to a guild hall and meet up with a player from bob.<mmo-name-here>.wtf before heading off for a quest.

Finally, it allows the community to take control of managing the game for you. If you decide that you need to move on to other projects and shut the game servers down, the still-dedicated fanbase can take over, moderating and maintaining the network after you've moved on to other things as a developer. Never again would someone have to see their favorite game become a shiny plastic paperweight (or more likely, a clump of useless bits), and this I see as the greatest incentive to switch to such a system.

The only technical obstacle in the face of this is speed. Unlike IRC, which is designed to transmit text data (which is easy), this protocol would be transmitting binary game data between multiple servers and clients. How would we quickly send the data packets from client to client if they're connected to servers across the world? One part of the solution is which transport-layer protocol we use- TCP or UDP. The other part (or parts) will be up to the designer of the protocol.

If anybody has any ideas, comments, or feedback to contribute, please feel free to do so.