[1] https://docs.freenet.org/tutorial.html
The fundamental problem is that the web is client-server, which inherently leads to concentration of power - and this is what we've seen over the past few decades.
Freenet replaces the client-server web with something entirely decentralized. More detailed explanation at [1].
Our approach to resource usage is that peers keep track of the resources used by each contract or on behalf of each neighboring peer, and compares it to the value provided by the contract or neighboring peer to our peer or the network. Peers or contracts with the least bang for the buck will get dropped by the peer when it is resource constrained. See [1] for how this works with contracts, and [2] for implementation.
[1] https://github.com/freenet/locutus/issues/244
[2] https://github.com/freenet/locutus/blob/3e5a31780f235e04f7b2...
See the diagram here: [1]
> From my perspective as a user, if after doing the drop in replacement, I can't use my bank's web site or browse Hacker News the same way, or watch TV on Hulu, it's not a drop in replacement because what I did before no longer works
Those are centralized services, the idea is to allow the creation of decentralized alternatives that can still be used through familiar tools like the web browser.
Many things about the new Freenet are very different, however the original Freenet had a reputation system called "web-of-trust"[1]. The new Freenet's reputation system will be a generalization of this.
> where you initially buy keywords from freenet.org, like a domain name purchase, and then other users add or subtract from your rating.
Not exactly buying keywords, you can use a donation to "bootstrap" a reputation, this is done anonymously through an RSA blind signature. There will be other ways to do this too - but a donation is simple, and obviously benefits the project.
> The new reputation mechanism must integrate with the subscription multicast in order to limit the abuse of the network. I'd be interested in a description on how this works.
This is largely addressed by lower-level mechanisms although the reputation system could also play a role. Each peer keeps track of its neighboring peer's behavior, abusive peers will be disconnected over time. See [2] for a description of part of this mechanism.
[1] https://freesocial.draketo.de/wot_en.html
[2] https://github.com/freenet/locutus/blob/3e5a31780f235e04f7b2...
> Can you explain the reputation system a bit more?
While it's still at the conceptual stage, the reputation system would be a generalization of a system called "web of trust" in the original Freenet [1].
> Is all content in freenet public or how would you know that someone posted certain content?
It depends on the system. For example, something like a forum might have minimum reputation requirements for posting, and also the requirement that the forum can downvote your reputation if you behave badly - this being the punishment.
> Do other users actively have to "down vote" them to reduce their reputation or is there an entity which can influence reputation?
It's a decentralized mechanism so it's all based on user feedback, but with web of trust it's transitive, so downvoting you may also downvote those who endorsed your reputation.
But bots can also participate, for example a bot could use modern deep learning to identify objectionable material and downvote automatically.
> Is there are single global reputation system or is it more local?
It's a mixture - trust is transitive so proximity matters but it isn't just about your immediate trust relationships.