|
· Föderation EN Di 21.01.2025 19:21:58 @katharina_buholzer @kubikpixel the relay currently is a centralized component (because only Bluesky (the company) provided one which covers all the PDS). But ATproto allows to have several. |
|
Föderation EN Di 21.01.2025 19:31:22 @nohillside @katharina_buholzer I am referring to this article, among other things, and it is not diletant: |
|
Föderation EN Di 21.01.2025 19:34:08 @kubikpixel @katharina_buholzer I know, I wrote one myself :-) There are currently two central elements, the relay and the DID generator. Both are implementation-side problems, not inherent to the protocol. |
|
Föderation EN Di 21.01.2025 19:36:38 @nohillside @kubikpixel Sure. I know that. And building a reliable broker involves access to considerable computing power. My point is: the entire activity pub architecture (fediverse etc) was build by a freshman as a semesterarbeit and does not really contemplate what distributed computing involves. That is why the background services on a daily basis crash with http 503 or have ddos like phenomena. I really know what I talk about here. |
|
Föderation EN Mi 22.01.2025 08:20:40 @katharina_buholzer @kubikpixel arenβt you mixing architecture with implementation here? The ActivityPub protocol itself doesnβt suffer from 503 ;-) Anyway: both protocols struggle with scaling, just on different aspects. Which kind of is unavoidable, scaling decentralized systems is hard. |
