> Doesn't this prove the value to users of a centralized platform over a federated platform?
Not at all. These types of centralized platforms won because they raised capital to grow fast and meet market needs before anything else. Email is still king, but if it were to be designed today, it would have been designed as some proprietary centralized system.
Distributed systems are hard.
In any case, I don't think federation goes far enough to save us from the problems we're seeing. I want a peer-to-peer social feed / social news app that doesn't depend on federation or servers at all.
Takedowns (DeCSS), takeovers (Twitter, FreeNode), shutdowns (Google+, Orkut, Digg, mastodon.technology), censorship (everywhere; should be an individual choice), maintainer-imposed spying (Apple CSAM), and maintainer-imposed changes or limits (Digg, modern Reddit, Twitter API) seem an order of magnitude harder to pull off if we have full control at the protocol and node level.
Federation would be nice for anonymity and aggregation of interest graph metadata, but at the core we should just have a swarm of content to sample and consume. It's fine if the content is naturally ephemeral as a consequence. We can use a constellation of opt-in 3rd party distributed (federated) services to provide durability, ranking, recommendation, filtering, etc. where desired.
Bittorrent, but for Twitter/Reddit/HN.