this post was submitted on 15 Apr 2025
1521 points (98.4% liked)
Technology
69098 readers
4321 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Several people have self hosted relays. Afaik nothing that anyone has used in "production", everyone just uses the default one. I expect that will change as people figure it out, and trust in bsky pbc drops with things like the current Turkish censorship incident
Example of self hosting https://bsky.app/profile/why.bsky.team/post/3lkwg2djrfk23
The code to run a relay is here https://github.com/bluesky-social/indigo
Your "example of self hosting" is not an example of self hosting the relay, just an appview which is still being fully dependent of other Bluesky services like the relay. It's pretty unlikely that the relay would be at all practical to host on a RPi5. But even if it was the problem still remains that the network is set up in a way where self-hosting it only results in you creating your own separate bubble, not meaningfully participating in the official one.
I also doubt anyone has selfhosted relays long-term since right now there's very little purpose to that and the resource requirements are massive as well as keep growing at a fast pace in terms of the disk space required.
The whole architecture is built around content addressing and allowing every account hosting server (PDS) talk to multiple relays and to allowing mirroring.
The whole point is to NOT create bubbles.
People already run their own PDS servers and participate with the official bluesky network, and can talk to users there, because their self hosted PDS syncs to the bluesky relay.
If you run your own relay and appview it STILL works, and you can talk without bubbles, if you still link your PDS to the bluesky relay to make yourself visible to their users, and if you set your appview / relay to retrieve content from the bluesky relay then you see content from bluesky users too.
Self hosted relays do exist, they're just not open to the public (mostly used for archival / development currently)
PDS is not very significant, it's just a tiny piece of the puzzle and doesn't really prove anything about the architecture. See this for more on what I'm getting at: https://neuromatch.social/@jonny/113365406995624763
That post is very misguided.
First of all, he's saying "you SHOULD make your PDS invisible to the bluesky servers because otherwise what's the point", but that's exactly equivalent to saying "our community want it's own Mastodon server - that means we MUST defederate Mastodon.social or what's the point?"
That's nonsense. Don't enforce silos on people.
Also, which relays to support are not chosen by users, it's chosen by the services the users choose. The PDS choose which relays to sync to, the appview does too, just like feed generators and moderation labelers does. They can trivially sync to multiple.
What people will choose is what app to use. This app will choose default appviews, and that appview chooses a relay, etc. Then they register an account, and the app suggests a default PDS server, or they host their own.
Also moderation labelers can be shared. Communities can run their own, and different communities who trust each other can import labels generated by the others.
Hosting a PDS is very cheap, it's just storage and bandwidth for the posts multiplied by the number of relays you directly sync to. With a few users on each that's nothing. It's in the range of free tier VPS hosting, RPi grade.
Deduplicating is probably the most trivial part. There's already code for handling duplicate events in streams. But more practically speaking, there's algorithms like set reconciliation which can make it significantly more bandwidth efficient to subscribe to multiple relays even when they have overlapping content.
Tldr no there won't be bubbles, unless that's what the users want. They surely CAN choose services which filter out the bluesky servers and which don't make them visible to bluesky, but why?
As for the DID part, the alternative is DID:Web, which requires permanent control over your domain name. With DID:PLC you can control your data by registering your own keys, although there's possibility for censorship. Their goal is to separate out running the DID:PLC service to an independent foundation.
As for the followup comments, just a few months ago bluesky made it significantly cheaper to authenticate jetstream events via Merkle tree diffs (jetstream is the lower bandwidth version of the relay firehose service). This means you can verify correctness quickly just by having a copy of the Merkle root hash & pubkey for the accounts you're interested in, you don't need to store the whole user repositories (excellent for feed generators and moderation labelers and anybody else doing partial sync)
I don't think you got the point tbh. It isn't about wanting to separate but about how dependent you are on Bluesky Corp. in every other scenario (and how hard it would be to deal with the situation if they decide to go rogue).
But that IS the point. The possibility of running independently PLUS the ability of bluesky users to migrate their account wholesale away from bluesky servers to 3rd party servers means you're not dependent on them.
They're literally designing for the principle of "the company is a future adversary" (see: Twitter, et al).
Yes and the thread I linked to explained why it is not looking like it's particularly well thought out for that case. Even beyond those issues they've always seemed very naive about what the company turning adversarial would actually be able to do but then again they obviously also have to worry about making money.