Bluesky is more open than you think.
-
I see a lot of misinformation about bluesky here, so I want to address a lot of the talking points against atproto/bluesky.
This is partially inspired by accounts like mastodon migration and feditips being really annoying about bluesky.
How Bluesky Works
I see a lot of people misunderstanding how it works.
The network has three main parts:- A PDS -- This stands for Personal Data Server. These store information in records, like who you are following, your posts, who you are blocking and your images.
- A relay -- These crawl PDSes and keep a copy of all the records on them. They give a "Firehose" of all the data on the network (that they crawled).
- An AppView -- These index and work through the data from the firehose. All interactions are handled through these, meaning if someone follows me on bluesky, that
app.bsky.graph.followrecord will be crawled by the relay, and recieved by the AppView. https://bsky.app/ is an Appview. Appviews don't always have to use the relays, https://whtwnd.com/ connects to PDSes directly.
This is different to ActivityPub, where if I follow someone, my server sends that information directly to the other person's server.
Common misconceptions
An atproto relay is too expensive to run.
https://atproto.africa/ is a second full-network relay run by the blacksky team. We already have a second relay, and they're not even that expensive to run anymore, a lot of people run non-archival (meaning it doesn't backfill every post) relays for less than $40 a month.
There is no instances available except for bsky.social
bsky.social isn't actually an instance, its just the domain name assigned to users by default. This is explained here: https://app.wafrn.net/fediverse/post/f8fc8da8-cd7e-4fae-a895-ac59dc28088f
Wafrn has (opt-in) bluesky support, they act as a PDS and AppView, so if bluesky disappears tomorrow they can switch to the atproto.africa relay. (There is DID:PLC which is a problem, but I'll get to that later.)
You can't defederate bsky.social, this proves atproto is centralised!
https://app.wafrn.net/fediverse/post/f8fc8da8-cd7e-4fae-a895-ac59dc28088f also explains this, bsky.social is just the name assigned to users, each PDS has names like https://brittlegill.us-west.host.bsky.network/ (where my account is).
While you could ignore records from a specific PDS on the App layer, its pretty pointless, since atproto is portable/content addressed, meaning a user could seamlessly move to another PDS. (AP does support moving, but its pretty seamful.)
(While I was writing this someone posted a pretty good blogpost about this: https://blog.cyrneko.eu/there-is-no-bsky-social-instance)
Bluesky can censor people in turkey, this proves they're centralised!
Those posts weren't removed, people on third party bluesky apps in turkey could still see them.
People in Turkey are automatically subscribed to a Moderation Service which hides those posts, as the government requires it.
If a person unsubscribes, or uses a third party app/server the posts are still there.Bluesky isn't decentralised as someone was banned for pointing out the head of T&S liked jailbait porn.
That person came back on a different PDS. They literally are still on bluesky because they joined a different server.
Bluesky went down due to a DDoS, this proves they are centralised!
The DDoS only crashed the Bluesky PDSes. People self hosting were fine.
Wafrn
Wafrn is a federated tumblr alternative. It started off as a tumblr clone, the dev added AP support, and eventually, Atproto support.
Its a great example of how bluesky can be built on.
If bluesky disappeared tomorrow, Wafrn could switch relays to atproto.africa, and still interact with people on other PDSes.
AppViewLite
appviewlite is a cool project I forgot to mention in the original post. It lets you self host an extremely lightweight Appview.
You can crawl PDSes yourself, eliminating the need for a relay.
https://github.com/alnkesq/AppViewLiteThe main reason I made this post is because so many people are blindly anti-atproto, without fully understanding how it works and how it can be improved.
There is obviously problems with it, but it does a lot right. (There's a lot ActivityPub should do, like content addressing, DIDs and composable moderation).
I also think we could do with a better bridge. bridgy isn't really cutting it right now.
Note on did:plc, its the only centralised part of the network as of now, its essentially the underlying ID every account has. It is possible to use a did:web id instead, which is tied to a website name.
PDSes and relays exist at the whim of Bluesky's corporate entity. Having all of the endpoints on the network controlled by a single agent is what makes Bluesky centralized. If Bluesky decided so, your server can be removed from their network and is functionally useless at that point. They decide who is and is not allowed to be a part of Bluesky.
For contrast, no such governing body exists with ActivityPub networks. Nobody can decide whether or not an instance should be removed from the network, they can only choose whether or not to federate with that instance. If you wanted to truly silence a Lemmy instance, for example, it would take the cooperation of all the major Lemmy admins to defederate, and is an entirely democratic process as a result.
EDIT: To clarify, ATProto is not what is centralized, "Bluesky" the platform utilizing ATProto, is what's centralized.
-
PDSes and relays exist at the whim of Bluesky's corporate entity. Having all of the endpoints on the network controlled by a single agent is what makes Bluesky centralized. If Bluesky decided so, your server can be removed from their network and is functionally useless at that point. They decide who is and is not allowed to be a part of Bluesky.
For contrast, no such governing body exists with ActivityPub networks. Nobody can decide whether or not an instance should be removed from the network, they can only choose whether or not to federate with that instance. If you wanted to truly silence a Lemmy instance, for example, it would take the cooperation of all the major Lemmy admins to defederate, and is an entirely democratic process as a result.
EDIT: To clarify, ATProto is not what is centralized, "Bluesky" the platform utilizing ATProto, is what's centralized.
wrote last edited by [email protected]Well, that problem also exists with mastodon.social and a lot of the actual fediverse.
Its less decentralised, but its still distributed. -
Well, if you have read it then your cherry picking of minor ways Bluesky is slighly less closed comes accross as pretty bad faith
I have read it, and a lot of the problems have been addressed.
Bluesky is still very early. There was an awkward period where lemmy was mostly lemmy.ml.I'm not trying to cherry pick anything, I'm just addressing arguments against it I have seen.
-
I see a lot of misinformation about bluesky here, so I want to address a lot of the talking points against atproto/bluesky.
This is partially inspired by accounts like mastodon migration and feditips being really annoying about bluesky.
How Bluesky Works
I see a lot of people misunderstanding how it works.
The network has three main parts:- A PDS -- This stands for Personal Data Server. These store information in records, like who you are following, your posts, who you are blocking and your images.
- A relay -- These crawl PDSes and keep a copy of all the records on them. They give a "Firehose" of all the data on the network (that they crawled).
- An AppView -- These index and work through the data from the firehose. All interactions are handled through these, meaning if someone follows me on bluesky, that
app.bsky.graph.followrecord will be crawled by the relay, and recieved by the AppView. https://bsky.app/ is an Appview. Appviews don't always have to use the relays, https://whtwnd.com/ connects to PDSes directly.
This is different to ActivityPub, where if I follow someone, my server sends that information directly to the other person's server.
Common misconceptions
An atproto relay is too expensive to run.
https://atproto.africa/ is a second full-network relay run by the blacksky team. We already have a second relay, and they're not even that expensive to run anymore, a lot of people run non-archival (meaning it doesn't backfill every post) relays for less than $40 a month.
There is no instances available except for bsky.social
bsky.social isn't actually an instance, its just the domain name assigned to users by default. This is explained here: https://app.wafrn.net/fediverse/post/f8fc8da8-cd7e-4fae-a895-ac59dc28088f
Wafrn has (opt-in) bluesky support, they act as a PDS and AppView, so if bluesky disappears tomorrow they can switch to the atproto.africa relay. (There is DID:PLC which is a problem, but I'll get to that later.)
You can't defederate bsky.social, this proves atproto is centralised!
https://app.wafrn.net/fediverse/post/f8fc8da8-cd7e-4fae-a895-ac59dc28088f also explains this, bsky.social is just the name assigned to users, each PDS has names like https://brittlegill.us-west.host.bsky.network/ (where my account is).
While you could ignore records from a specific PDS on the App layer, its pretty pointless, since atproto is portable/content addressed, meaning a user could seamlessly move to another PDS. (AP does support moving, but its pretty seamful.)
(While I was writing this someone posted a pretty good blogpost about this: https://blog.cyrneko.eu/there-is-no-bsky-social-instance)
Bluesky can censor people in turkey, this proves they're centralised!
Those posts weren't removed, people on third party bluesky apps in turkey could still see them.
People in Turkey are automatically subscribed to a Moderation Service which hides those posts, as the government requires it.
If a person unsubscribes, or uses a third party app/server the posts are still there.Bluesky isn't decentralised as someone was banned for pointing out the head of T&S liked jailbait porn.
That person came back on a different PDS. They literally are still on bluesky because they joined a different server.
Bluesky went down due to a DDoS, this proves they are centralised!
The DDoS only crashed the Bluesky PDSes. People self hosting were fine.
Wafrn
Wafrn is a federated tumblr alternative. It started off as a tumblr clone, the dev added AP support, and eventually, Atproto support.
Its a great example of how bluesky can be built on.
If bluesky disappeared tomorrow, Wafrn could switch relays to atproto.africa, and still interact with people on other PDSes.
AppViewLite
appviewlite is a cool project I forgot to mention in the original post. It lets you self host an extremely lightweight Appview.
You can crawl PDSes yourself, eliminating the need for a relay.
https://github.com/alnkesq/AppViewLiteThe main reason I made this post is because so many people are blindly anti-atproto, without fully understanding how it works and how it can be improved.
There is obviously problems with it, but it does a lot right. (There's a lot ActivityPub should do, like content addressing, DIDs and composable moderation).
I also think we could do with a better bridge. bridgy isn't really cutting it right now.
Note on did:plc, its the only centralised part of the network as of now, its essentially the underlying ID every account has. It is possible to use a did:web id instead, which is tied to a website name.
As I understand (I could be wrong) bridgy is not useful as it could be as it got bullied into being opt-in instead of opt-out.
-
I have read it, and a lot of the problems have been addressed.
Bluesky is still very early. There was an awkward period where lemmy was mostly lemmy.ml.I'm not trying to cherry pick anything, I'm just addressing arguments against it I have seen.
Lemmy.ml, lemmy.world, lemmy.zip and any other instance run on the same software
Wafrn doesn't run the same software as Bluesky.social
-
I see a lot of misinformation about bluesky here, so I want to address a lot of the talking points against atproto/bluesky.
This is partially inspired by accounts like mastodon migration and feditips being really annoying about bluesky.
How Bluesky Works
I see a lot of people misunderstanding how it works.
The network has three main parts:- A PDS -- This stands for Personal Data Server. These store information in records, like who you are following, your posts, who you are blocking and your images.
- A relay -- These crawl PDSes and keep a copy of all the records on them. They give a "Firehose" of all the data on the network (that they crawled).
- An AppView -- These index and work through the data from the firehose. All interactions are handled through these, meaning if someone follows me on bluesky, that
app.bsky.graph.followrecord will be crawled by the relay, and recieved by the AppView. https://bsky.app/ is an Appview. Appviews don't always have to use the relays, https://whtwnd.com/ connects to PDSes directly.
This is different to ActivityPub, where if I follow someone, my server sends that information directly to the other person's server.
Common misconceptions
An atproto relay is too expensive to run.
https://atproto.africa/ is a second full-network relay run by the blacksky team. We already have a second relay, and they're not even that expensive to run anymore, a lot of people run non-archival (meaning it doesn't backfill every post) relays for less than $40 a month.
There is no instances available except for bsky.social
bsky.social isn't actually an instance, its just the domain name assigned to users by default. This is explained here: https://app.wafrn.net/fediverse/post/f8fc8da8-cd7e-4fae-a895-ac59dc28088f
Wafrn has (opt-in) bluesky support, they act as a PDS and AppView, so if bluesky disappears tomorrow they can switch to the atproto.africa relay. (There is DID:PLC which is a problem, but I'll get to that later.)
You can't defederate bsky.social, this proves atproto is centralised!
https://app.wafrn.net/fediverse/post/f8fc8da8-cd7e-4fae-a895-ac59dc28088f also explains this, bsky.social is just the name assigned to users, each PDS has names like https://brittlegill.us-west.host.bsky.network/ (where my account is).
While you could ignore records from a specific PDS on the App layer, its pretty pointless, since atproto is portable/content addressed, meaning a user could seamlessly move to another PDS. (AP does support moving, but its pretty seamful.)
(While I was writing this someone posted a pretty good blogpost about this: https://blog.cyrneko.eu/there-is-no-bsky-social-instance)
Bluesky can censor people in turkey, this proves they're centralised!
Those posts weren't removed, people on third party bluesky apps in turkey could still see them.
People in Turkey are automatically subscribed to a Moderation Service which hides those posts, as the government requires it.
If a person unsubscribes, or uses a third party app/server the posts are still there.Bluesky isn't decentralised as someone was banned for pointing out the head of T&S liked jailbait porn.
That person came back on a different PDS. They literally are still on bluesky because they joined a different server.
Bluesky went down due to a DDoS, this proves they are centralised!
The DDoS only crashed the Bluesky PDSes. People self hosting were fine.
Wafrn
Wafrn is a federated tumblr alternative. It started off as a tumblr clone, the dev added AP support, and eventually, Atproto support.
Its a great example of how bluesky can be built on.
If bluesky disappeared tomorrow, Wafrn could switch relays to atproto.africa, and still interact with people on other PDSes.
AppViewLite
appviewlite is a cool project I forgot to mention in the original post. It lets you self host an extremely lightweight Appview.
You can crawl PDSes yourself, eliminating the need for a relay.
https://github.com/alnkesq/AppViewLiteThe main reason I made this post is because so many people are blindly anti-atproto, without fully understanding how it works and how it can be improved.
There is obviously problems with it, but it does a lot right. (There's a lot ActivityPub should do, like content addressing, DIDs and composable moderation).
I also think we could do with a better bridge. bridgy isn't really cutting it right now.
Note on did:plc, its the only centralised part of the network as of now, its essentially the underlying ID every account has. It is possible to use a did:web id instead, which is tied to a website name.
I hope I am not adding to the problem here as well. It seems that obviously Bluesky is neither fully centralized nor fully decentralized. Is there a statement about just how much of either it is?
Although that might be complicated - like someone could say that Lemmy is fairly centralized, bc if you block Lemmy.World then you lose half the users and perhaps half the communities (and PieFed even more so, with PieFed.social representing an even higher fraction of users and communities on it).
So there is a distinction between Bluesky the service as it currently is implemented and Bluesky the protocol, the former of which is fairly centralized but the latter is more expandable?
-
I hope I am not adding to the problem here as well. It seems that obviously Bluesky is neither fully centralized nor fully decentralized. Is there a statement about just how much of either it is?
Although that might be complicated - like someone could say that Lemmy is fairly centralized, bc if you block Lemmy.World then you lose half the users and perhaps half the communities (and PieFed even more so, with PieFed.social representing an even higher fraction of users and communities on it).
So there is a distinction between Bluesky the service as it currently is implemented and Bluesky the protocol, the former of which is fairly centralized but the latter is more expandable?
if you block Lemmy.World then you lose half the users
35% (16k out of 46k MAU): https://lemmy.fediverse.observer/list
-
if you block Lemmy.World then you lose half the users
35% (16k out of 46k MAU): https://lemmy.fediverse.observer/list
Nice. I remember when it was 80%, then it fell to half, 40%, and apparently now is closer to a third than half. Excellent decentralizing!:-)
-
if you block Lemmy.World then you lose half the users
35% (16k out of 46k MAU): https://lemmy.fediverse.observer/list
Or even 33% as we should count PieFed and Mbin too (this makes 48k MAU overall). All 3 "apps" make one network.
-
Or even 33% as we should count PieFed and Mbin too (this makes 48k MAU overall). All 3 "apps" make one network.
Good point!
-
Very useful, thanks.
As I see it, Bluesky is fundamentally different from Xitter and it is a major step in the right direction. It is short-sighted to reject it because of some technical imperfections.
The fundamental question IMO is whether there is enough mindshare (i.e. users and attention) to allow ATSocial (AKA partial federation) and ActivityPub (AKA total federation) to both be successful. I'm thinking there is. After all, the vast majority of people are still on ad-fuelled corporate social media, with all its internal contradictions.
I think the technical imperfections are not the real reason people are against it. In my opinion it just can't be trusted to have a corp in control. It would be like having Microsoft own the activity pub repo.
-
I hope I am not adding to the problem here as well. It seems that obviously Bluesky is neither fully centralized nor fully decentralized. Is there a statement about just how much of either it is?
Although that might be complicated - like someone could say that Lemmy is fairly centralized, bc if you block Lemmy.World then you lose half the users and perhaps half the communities (and PieFed even more so, with PieFed.social representing an even higher fraction of users and communities on it).
So there is a distinction between Bluesky the service as it currently is implemented and Bluesky the protocol, the former of which is fairly centralized but the latter is more expandable?
I just say bluesky because that's what everyone knows it as. I'm really talking about its network.
Its not very well distributed, because almost everyone is on bluesky's meganodes.
Its more of a social problem than a technical problem at this point. -
Lemmy.ml, lemmy.world, lemmy.zip and any other instance run on the same software
Wafrn doesn't run the same software as Bluesky.social
My point was that the network was fairly centralised in the beginning. The people behind atproto.africa are working on an alternate bluesky appview anyway.
-
As I understand (I could be wrong) bridgy is not useful as it could be as it got bullied into being opt-in instead of opt-out.
You would be correct.
-
I think the technical imperfections are not the real reason people are against it. In my opinion it just can't be trusted to have a corp in control. It would be like having Microsoft own the activity pub repo.
I agree with you there.
I wish they put a bit more effort into getting people onto independant servers.
They took to opposite approch of mastodon: they abandoned proper distribution for better growth.In any case, ActivityPub and atproto can both coexist.
-
My point was that the network was fairly centralised in the beginning. The people behind atproto.africa are working on an alternate bluesky appview anyway.
Good to hear!
The main difference is still that every work put into Bluesky.social can not be reused by other "servers", unlike Lemmy
-
My point was that the network was fairly centralised in the beginning. The people behind atproto.africa are working on an alternate bluesky appview anyway.
Still is. Always will be.
-
Well, that problem also exists with mastodon.social and a lot of the actual fediverse.
Its less decentralised, but its still distributed.Mastodon is more open than you think.
You made a post to attempt to dispell what you consider a misunderstanding about BlueSky, yet your comment suggests you dont understand the Fediverse.
Well, that problem also exists with mastodon.social
No, it doesn't. There are thousands of instances, some with hundreds of thousands of users. If you sort the instance list by active users, the population spreads out even more, because smaller instances have more active users.
and a lot of the actual fediverse.
Wrong again. Lemmy.world is about 30% of Lemmy, and less when you include Mbin, PieFed, etc.
Its less distributed, but its still decentralised.
I run a Fedi instance connected to hundreds of others. If one, even a large one, defederates me, it does not cut me off. If I ran a PDS, I'd be connected to BlueSky, and they can do what they want.
You are rationalizing this to yourself because you like BlueSky.
-
Well, that problem also exists with mastodon.social and a lot of the actual fediverse.
Its less decentralised, but its still distributed.That's at a very different level. With dot social it's about a quarter of the active users on the fediverse, whereas bluesky is probably something like 95% centralized in practice. It seems to keep improving, but right now it's basically impossible to use without mostly interacting with bsky.
-
Mastodon is more open than you think.
You made a post to attempt to dispell what you consider a misunderstanding about BlueSky, yet your comment suggests you dont understand the Fediverse.
Well, that problem also exists with mastodon.social
No, it doesn't. There are thousands of instances, some with hundreds of thousands of users. If you sort the instance list by active users, the population spreads out even more, because smaller instances have more active users.
and a lot of the actual fediverse.
Wrong again. Lemmy.world is about 30% of Lemmy, and less when you include Mbin, PieFed, etc.
Its less distributed, but its still decentralised.
I run a Fedi instance connected to hundreds of others. If one, even a large one, defederates me, it does not cut me off. If I ran a PDS, I'd be connected to BlueSky, and they can do what they want.
You are rationalizing this to yourself because you like BlueSky.
You're misinterpreting my comment, I said that getting cut off a large server is a problem in any network. The problem is worse on bluesky, but that can change.
If I run a PDS, I connect to bluesky and other instances as well. I can get cut off bluesky's server, but there is other servers.I'm not rationalising anything, I just think the discourse around bluesky is toxic, and I want to at least make it less annoying.
I don't even like bluesky, I do like the underlying protocol though.
