Post #2346342

2026-05-07 10:34 UTC

@strypey@mastodon.nzoss.nz @uriel@x.keinpfusch.net I can see why "compatibility first" in your Aktor project is giving you frustrations when finding out what happens on the wire. 😅 The app-centric post-facto interop work method that prevails has a whole range of negative side effects. Everyone who creates an arbitrary AP extension becomes owner as it were of that part of the spec. And if the spec now works for their app they are no longer that incentivised to improve it, or even actively resist change, as that would be extra work. And the next person to adopt some existing AP extension is the technology decision maker to take on a foreign dependency, not unlike JS/TS NPM dependency hell world. But on protocol level. What we currently see is that precious few people really care about the ActivityPub protocol wrt its beneficial role for the ecosystem as a whole. People care to the extent it serves their own app. And for the current popular app platforms it does it well enough, and the devs mastered the complexity.

Replies (2)

@smallcircles@social.coop 2026-05-07 10:44

@strypey@mastodon.nzoss.nz @uriel@x.keinpfusch.net Oh btw, there are a couple of very good IETF documents on protocol design. The term "protocol decay" comes from one of those. See: https://coding.social/blog/grassroots-evolution/#ietf-best-practices

Open ##2346343
@uriel@x.keinpfusch.net 2026-05-07 11:02

@smallcircles@social.coop @strypey@mastodon.nzoss.nz Honestly, I was mostly thinking about the discussions around Mastodon when it transitioned from 32-bit to 64-bit IDs. That said, your point does make sense to me. If, in the future, people start implementing new extensions — groups, marketplaces, or other features — they may gradually end up evolving in somewhat separate ecosystems. I think this is also part of the reason why projects such as Lemmy , federated marketplace experiments, or even Mobilizon can be difficult to integrate cleanly with the wider network: each tends to develop its own assumptions, semantics, and workflows. So I can see the argument for stronger capability discovery mechanisms, whether at the instance level or, in some cases, even at the individual user level.

Open ##2346344