Decentralizing Wikipedia and its sister projects

Decentralizing Wikipedia and its sister projects is a project to formulate a framework of how to decentralize the subsystems of the various Wikimedia projects.

In practice, this project could perhaps be a fork of BOINC, GnuTella, OFFsystem or FreeNet or some synthesis of the aforementioned... or something else.

Brainstorming
It might work something like the following. Anyone can setup tracker and/or content server (that runs as part of Wikiversity Screen Saver, some other program, or just as a service in the background). These servers and nodes track changes and where you can find articles on the network.

Trackers could be large or small. They could be dedicated machines, or run in the background of peoples' electronic devices.

For example, since Wikipedia is quite large, Wikipedia would probably have the most tracker servers.

Wikiversity would probably have a great deal fewer. All of this could be dynamically balanced and tweaked according to need.

When someone wants to view an article on cows, they basically ping a tracker server, saying in computer talk, "I want to see an article on Cows".

Their machine could attempt to tell this to multiple tracker servers, and could hopefully get a rather speedy response.

Trust
The tracker server would say, "In order the amount of trust you can put into the integrity of the article, you can find the most up to date revision of the article on Cows on the servers x, y, z, and c".

The person who requested the article on cows would then send a message off to machines x, y, z, and c". Hopefully, x would respond first, in a speedy manner and the article would pop up speedily.

Presuming this goes right, the individual may decide he or she would like to edit the article.

They make their changes and send them off to a few of the servers that contain the article on cows. This would then be propagated and perhaps reviewed by editors and then integrated into the best version.

Maybe it could simply be integrated. Suppose someone came along and saw that it was vandalism. They could revert it as normally is done, and then that revert could be propagated.

Maybe OpenID could be used to verify identities. That is an interesting challenge. How could identities be trusted on a decentralized system. OpenId?

There could be separate nodes for images.

Questions

 * Would a whole copy of the all the text be needed on servers?
 * Could images and other media be stored in a more decentralized fashion?