The idea is essentially to follow the way email (and partly the Web) works: users aren’t tied to one server (centralized), but neither are they required to each run their own one (peer-to-peer), instead there’s a narrower group of server operators who host users, but those users can cause the server software to connect to a server it doesn’t know about if they mention it explicitly.
Of course, if the protocol is too weak in allowing the operator to control those connections (in either direction), it will evolve informal means for that which will make all but the largest servers extremely unreliable, much as email did. On the other hand, the experience of Mastodon shows the dangers of operators exercising too much control (where e.g. Mozilla seems eager to defederate with any server that has anyone post anything even slightly offensive or objectionable to anybody whose opinion Mozilla considers valid).
The federated wiki idea as promoted by Ward seems to be to have the federation (network of servers) be able to browse each other’s pages—so far so Web—but then to allow each user to clone and edit any page anywhere, storing the clone on their own server. The original page isn’t affected, except I think there’s a provision for some sort of backlinking (referral spam? what’s that?). It doesn’t sound unreasonable, but I’m not sure it can support anything interesting either—for a large pool of collaborators you’d need a Torvalds-like full-time merge BDFL, and I haven’t even seen a discussion of pull requests or anything similar.
The idea is essentially to follow the way email (and partly the Web) works: users aren’t tied to one server (centralized), but neither are they required to each run their own one (peer-to-peer), instead there’s a narrower group of server operators who host users, but those users can cause the server software to connect to a server it doesn’t know about if they mention it explicitly.
Of course, if the protocol is too weak in allowing the operator to control those connections (in either direction), it will evolve informal means for that which will make all but the largest servers extremely unreliable, much as email did. On the other hand, the experience of Mastodon shows the dangers of operators exercising too much control (where e.g. Mozilla seems eager to defederate with any server that has anyone post anything even slightly offensive or objectionable to anybody whose opinion Mozilla considers valid).
The federated wiki idea as promoted by Ward seems to be to have the federation (network of servers) be able to browse each other’s pages—so far so Web—but then to allow each user to clone and edit any page anywhere, storing the clone on their own server. The original page isn’t affected, except I think there’s a provision for some sort of backlinking (referral spam? what’s that?). It doesn’t sound unreasonable, but I’m not sure it can support anything interesting either—for a large pool of collaborators you’d need a Torvalds-like full-time merge BDFL, and I haven’t even seen a discussion of pull requests or anything similar.