because hosting a full-network relay is super cheap, including bandwidth. there are multiple people who are running full-network relays (monitoring and relaying everything from every PDS) for less than 30USD per month

https://fediversereport.com/atmosphere-report-116/

[–] kbal@fedia.io 2 points 4 days ago

Well that's an interesting development, which people seem to have begun experimenting with in the past few weeks. Of course for the time being I suppose it's made somewhat easier for the relays by 99% of the users being hosted on the small collection of official bsky.network PDS servers.

Some more discussion of it in which there may be hints as to how atproto people might slowly progress towards reinventing some of the things we take for granted on the fediverse.

[–] Blaze@lemmy.dbzer0.com 1 points 4 days ago (1 children)

This news probably deserve its own post

[–] wisdomchicken@piefed.social 1 points 3 days ago (1 children)

i think the main takeaway is that the fediverse has hugely overindexed on relays being this big huge centralising force in the atproto network. And thats simply not true at all. The flipside of that is that relays also dont really matter much either. All they do is simply aggregating from a distributed network of data storage into a single firehose. Its really cool that you can do that for super cheap. but its also just a small part of the entire network architecture. like, atproto relays are not CDNs, for example, and video CDNs are expensive to run.

[–] Blaze@lemmy.dbzer0.com 1 points 3 days ago (1 children)

Definitely. Posting https://whtwnd.com/bnewbold.net/3lo7a2a4qxg2l as its own post would probably help

[–] Irelephant@lemm.ee 1 points 3 days ago

The main centralised part of bluesky right now is their decentralised identifiers right now.

DID:PLC is completely centralised. Its not actually a requirement for a did to be decentralised.

plc.directory is the only registry for them right now.