ActivityPub Four Years Later: The Good, the Bad, and the Fedi
rolltime
Where
Talks
A New HOPE
Duration
47 min
Looks at how the ActivityPub federated standard has fared since 2018 and why it remains a niche alternative.
Summary
Rolltime, a long-time user rather than a spec author, revisits ActivityPub four years after it became a W3C standard. The talk starts by explaining federation: independent servers that pass messages on behalf of their users, sitting between fully centralized platforms and peer-to-peer designs that break when nodes go offline. It then traces the early fragmentation of federated systems, where incompatible protocols kept users apart, and how a single standard emerged to unify them.
The middle section looks at Mastodon as the largest platform in the fediverse. It covers chronological feeds without an algorithm, local and federated timelines, instance-level rules and blocking, content warnings, and alt text tooling. The speaker also points to gaps: the spec has no built-in identity or consent mechanism, implementations diverge on features such as follow requests so actions do not always reconcile, and most platforms are close clones of one another.
The closing argument is that adoption has stalled because timelines serve a consumption model shaped by corporate incentives. The speaker urges listeners to join the network, help improve the spec, and build interactive, creation-oriented social features instead of copies of existing products. Audience questions covered self-hosting, mobile push notifications, cross-posting bridges, and whether the protocol can succeed without mainstream adoption.
Related talks
Talks·56 min
Evan Prodromou
Traces how social networks evolved toward the Fediverse, a federated network where users interact across platforms.
Talks·32 min
Lee Felsenstein
Traces the free speech movement, early personal computing and community bulletin boards, and a proposed civic social media design.
Talks·38 min
Wesley Hales
Shows how inline classification of sensitive data in service proxies can flag data leaving a system in real time.


