This is technically incorrect (the best kind of incorrect?). Bluesky is open source, with the exception of the discover feed algorithm, which they claim must remain secret to prevent it being manipulated. There are open-source replacements for that feed available, so it’s open enough that it is theoretically possible to spin up a Bluesky replacement, albeit impossibly expensive.
Coming at it from another angle though, the product in any commercial social media product is you, so in that sense you’re right: the product is not open source. Either way, open source code is not some panacea that erases all risk of commodifying its users. Bluesky is a great example because while it is open source, that in absolutely no way prevents them from tracking their users.
It’s kind of complicated. Bluesky doesn’t do anything the way the fediverse does, so a PDS isn’t a full instance, it’s just the way that your personal account interacts with the Bluesky service.
An analogy I used in another thread about Bluesky got way too complicated, but my starting point was that if Bluesky is a swimming pool, then hosting a PDS is bringing your own personal bucket of water from home. Ultimately, you’re still feeding it all into the one big pool that Bluesky owns, at least until somebody else builds another swimming pool (puts up the money to host a fully-fledged Bluesky replacement service) and you take your bucket over there.
On its own, the PDS doesn’t really do anything without the rest of the infrastructure behind it. You can’t go swimming in a bucket.
Bluesky is a trailer park. A PDS is a trailer. You can take it somewhere else, but you need somewhere to park it at night, and right now the only option is Bluesky.
Yeah. Bluesky works way more like how people seem to imagine the fediverse does, with PDSes being glorified dumb terminals accessing a (functionally, if not forever technically) centealized pool of content. Hosting a PDS is just shouldering some of the cost of BS’s last mile.
This isn’t necessarily a bad thing though, it removed the complexity behind instances and federation which is primarily the reason mastodon didn’t see mass adoption.
It isn’t federation, since not all nodes in the system have equal power (control). There is still a central authority that controls what the inferior nodes can do.
Contrast that with email servers where you can send a message from one server to another without a more authoritative node as a required middleman.
Is there a list of all the relays and appview servers run by other people for the community to use? I looked for one, but could find no evidence that others have actually hosted instances of these components for real use.
I think the problem is there isn’t much incentive to run a relay in particular, as it all funnels into the same place, but if you’re questioning the tech this guy did it a few days ago with the relay
A pds can do a lot more than people suggest, but its not very effective.
Essentially, atproto has three distinct parts:
A PDS, stores your posts, user, and handles authentication
A relay, crawls every pds, and creates a “firehose” of data to build stuff with
An AppView, an app built with data from the relay. bsky.app is an appview, flushes.app is another. whtwnd.com is a blogging appview
The relay is the main “centralised” part of the network, but its possible to use the network without it. whtwnd, for example, crawls PDSs directly, without a relay.
Most likely. That’s if people knew about. You could do it secretly.
Though I wonder if you were open about it, if people would accept it. Just say “hey, this instance doesn’t ask for donations, but we track and sell your info.” Maybe some users would be okay with that.
they can’t sell your info (from remote instances) without you agreeing to a privacy policy. Now, that most likely wouldn’t stop them, but it makes it harder legally.
Remote instances won’t have your IP or email, and other usage trend data. So that info could only be obtained by this hypothetical tracking instance. As for any remote content on other instances, that can just be scraped by anyone. You wouldn’t even need an account or instance to get that data.
This is technically incorrect (the best kind of incorrect?). Bluesky is open source, with the exception of the discover feed algorithm, which they claim must remain secret to prevent it being manipulated. There are open-source replacements for that feed available, so it’s open enough that it is theoretically possible to spin up a Bluesky replacement, albeit impossibly expensive.
Coming at it from another angle though, the product in any commercial social media product is you, so in that sense you’re right: the product is not open source. Either way, open source code is not some panacea that erases all risk of commodifying its users. Bluesky is a great example because while it is open source, that in absolutely no way prevents them from tracking their users.
Additionally, it looks like you can host your own instance too.
PDS Self-hosting
It’s kind of complicated. Bluesky doesn’t do anything the way the fediverse does, so a PDS isn’t a full instance, it’s just the way that your personal account interacts with the Bluesky service.
An analogy I used in another thread about Bluesky got way too complicated, but my starting point was that if Bluesky is a swimming pool, then hosting a PDS is bringing your own personal bucket of water from home. Ultimately, you’re still feeding it all into the one big pool that Bluesky owns, at least until somebody else builds another swimming pool (puts up the money to host a fully-fledged Bluesky replacement service) and you take your bucket over there.
On its own, the PDS doesn’t really do anything without the rest of the infrastructure behind it. You can’t go swimming in a bucket.
Bluesky is a trailer park. A PDS is a trailer. You can take it somewhere else, but you need somewhere to park it at night, and right now the only option is Bluesky.
Yeah. Bluesky works way more like how people seem to imagine the fediverse does, with PDSes being glorified dumb terminals accessing a (functionally, if not forever technically) centealized pool of content. Hosting a PDS is just shouldering some of the cost of BS’s last mile.
not even: a copy of everything on your pds has to be replicated by bluesky’s relay.
This isn’t necessarily a bad thing though, it removed the complexity behind instances and federation which is primarily the reason mastodon didn’t see mass adoption.
It isn’t federation, since not all nodes in the system have equal power (control). There is still a central authority that controls what the inferior nodes can do.
Contrast that with email servers where you can send a message from one server to another without a more authoritative node as a required middleman.
Claiming it’s a central authority when anyone can run a relay is a little disingenuous.
Is there a list of all the relays and appview servers run by other people for the community to use? I looked for one, but could find no evidence that others have actually hosted instances of these components for real use.
Not sure if an absolute list, but https://atprotocol.dev/ is a good resource.
I think the problem is there isn’t much incentive to run a relay in particular, as it all funnels into the same place, but if you’re questioning the tech this guy did it a few days ago with the relay
https://whtwnd.com/futur.blue/3lkubavdilf2m
It looks like ^ he also has an appview up thats long form blogging here: https://github.com/whtwnd/whitewind-blog
Another guy also built a lightweight app view
https://bsky.app/profile/why.bsky.team/post/3lkwg2djrfk23
And stream.place is another appview but it’s not a microblog platform. Also very early stages I believe.
A pds can do a lot more than people suggest, but its not very effective.
Essentially, atproto has three distinct parts:
The relay is the main “centralised” part of the network, but its possible to use the network without it. whtwnd, for example, crawls PDSs directly, without a relay.
There’s more to it, but thats the basics.
There’s nothing to prevent someone from spinning up a lemmy or mastodon instance and tracking users either.
They’d get defederated quickly.
Most likely. That’s if people knew about. You could do it secretly.
Though I wonder if you were open about it, if people would accept it. Just say “hey, this instance doesn’t ask for donations, but we track and sell your info.” Maybe some users would be okay with that.
they can’t sell your info (from remote instances) without you agreeing to a privacy policy. Now, that most likely wouldn’t stop them, but it makes it harder legally.
Remote instances won’t have your IP or email, and other usage trend data. So that info could only be obtained by this hypothetical tracking instance. As for any remote content on other instances, that can just be scraped by anyone. You wouldn’t even need an account or instance to get that data.
I know, sure thats just what threads did.
Actually, you can host the App bit yourself, and piggyback off bluesky’s relay until another one pops up.