In my experience, the only acceptable way to deal with many classes of abuse is humans. You can cleave out a bunch of spam by making it harder / more expensive

In my experience, the only acceptable way to deal with many classes of abuse is humans. You can cleave out a bunch of spam by making it harder / more expensive / more annoying for spammers to operate, but I've never seen a community of humans that didn't require human labor in order to govern itself. Federated reputation is the right idea, but it's only a means. The real hard problem is distributing the mental / emotional labor of moderation. I don't think this decentralizes well. Good moderation requires focus, a carefully trained subjective sense, along pragmatic acquired experience for how to act on issues that come up. If you agree that the most efficient/appropriate route is communities having human moderators, then that puts a ton of constraints on any solution, and it begins to resemble the forum / IRC op / reddit moderator structures we've seen before. It's pretty simple to create a federated platform with a strong anti-abuse component, but the solution is not what most proponents of decentralized platforms want to hear. You need: * definitive scopes in which certain moderators have a ton of (possible unauditable!) power, * consensus on which moderators exist over which scopes, * encrypted group channels for communicating with moderators, * and, the most important part: a tremendous amount of dedicated human elbow-grease and emotional labor.