Thread

There’s a debate raging over in the Bluesky world about whether or not infrastructure providers on ATprotocol should be neutral carriers or if people running things like PDS servers should be able to choose who they host. It’s an interesting read and worth thinking about. From a Nostr perspective it’s like arguing for a custodial system then being upset at the power dynamics that exist because of that. I’m curious what folks think. I think the poster kicked a hornets nest, not understanding how communities of users react to being told what they should or shouldn’t do with their own servers. Thoughts?

Replies (71)

Infrastructure is just a service provided. Reminds me of a baker in the UK who refused to bake a cake for a gay couple based on their religious beliefs which created a lot of discussion. Should they be compelled to provide their service or have a right to choose? I think the only way to truly separate infrastructure from ideology is if the infrastructure (i.e. relays) have no view of what they are relaying. It’s probably not just going to be ideology either. There will be legislative reasons. Again, for the UK I wouldn’t be surprised if some relays services feel compelled to not offer services to UK residents. And we should make sure there’s more than 1 baker in town :-)
> "I think the only way to truly separate infrastructure from ideology is if the infrastructure (i.e. relays) have no view of what they are relaying." I completely agree, that *guarantees* neutrality. However, the entire internet and its neutrality today are not based on guarantees, but on culture. I expected AT Protocol to understand this, but they have turned it 180Β° and are making this cultural neutrality a threat. Thank you for the welcome :) I hope you'll forgive me for my shitposting about bitcoin sometimes ;)
Their position, broken down into consequences, is that culturally neutral infrastructure is the wrong approach. That we inherently support racists when we have neutral web hosts or email providers, that the right approach is ideology-based because it prioritizes "community safety". It seems a bit like the Core vs Knots case to me. One side wants to have nodes as neutral as possible and the other side want to bring unnecessary ideological baggage into it.
"I think the only way to truly separate infrastructure from ideology is if the infrastructure (i.e. relays) have no view of what they are relaying." I agree, but one could make the argument that as Nostr scales, this would be untenable because relays would be crushed by spam. But there are ways around that, I.e. Web of Trust and/or Proof of Work, that would obviate the necessity of analyzing the content of a message (and therefore being responsible for its transmission if it's "bad"), while still serving as an inhibitor of spam. In "high spam traffic", proof of work could even be set at (say) 1 second, which would be no problem for most users.
That’s where hashcash came from I believe and same principles here could be used here (using Bitcoin for relays rather than email) that it becomes uneconomical to post excessive spam. FYI @npub1qg8j...24kw That works for encrypted DM type notes. Not sure about public stuff though if the relay remained in the dark about what it’s transmitting unless the message is sharded and that sounds complicated.
Well you're not using "Bitcoin" per se, it's just adding proof of work to messages. Coracle does this. It just uses your computer to do proof of work on a hash of your message. No reason why it couldn't work for public notes too. Nothing is encrypted. It's the same kind of thing that Anubis does: Force the user instead of some "captcha" to do a small amount of computational work, that's no problem if it's a human but becomes uneconomical if it's a massive web scraper"
Yes I agree, It's the view of the masses, but also the majority opinion of the developers that make up the ecosystem (from my pov). I believe that Jay and the Bluesky team mean it honestly, but it's impossible for them to resist the community pressure and this "enshittification". Cory Doctorow diagnosed this a long time ago: https://doctorow.medium.com/https-pluralistic-net-2024-11-02-ulysses-pact-tie-yourself-to-a-federated-mast-b2f89bb5b4d8
A big part of these types of outcomes is also based on the community that forms around them. Bluesky ended up, for various reasons largely unrelated to its underlying protocol, being adopted by a very specific community, and that community views policing untoward speech as a core part of how social media should operate (not passing judgement on that here!). While the protocol design attempted to do that at a higher layer, the community’s strong stance on moderation and the developer’s need to fight urgent spam and moderation fires meant that protocol neutrality took a back seat. Once it was there, with a rapidly-growing community that viewed this as good, it wasn’t coming back.
Sure, but to borrow some ideas from #urbanism, I do think that "infrastructure determines culture" to a certain extent here.... The major barriers of entry for anyone like Blacksky to spin up alternative systems creates a situation where the "exit / voice" dynamic is heavily skewed towards "voice", and users feel their only option is to fight battles rather than do their own thing.
The only real way to make the relay not know what it's sending but not have everything encrypted is to make the operator unable to communicate preferences to the relay, which means relays in enclaves with attested runtime code meeting protocol spec, and if no proof-of-blind-enclave then not considered protocol compliant.
The enclave approach would be close to that, the operator could obviously know what's on the relay, because anyone can query the relay, and the notes are not encrypted. But the op wouldn't be able to do anything with this knowledge, because to change the code of the relay would require breaking the attestation and being booted from protocol compliance. So code-enforced not caring, in a way.
Thanks Rabble! I kicked that nest little on purpose. I was on atproto very early, among the first 40k people - and I chose atproto over Nostr because I thought it was a technology with greater scaling potential (not perfect, but which can bring freedom to more people). I saw atproto as the opposite approach to nostr, but converging to a very similar destination point. But the mood of the last few months clearly anticipated how wrong I was, that it had become just a decentralization theater. My assumption that the community would move towards decentralization was completely wrong and naive. The rage that has now finally confirmed my suspicions and given me the strength to leave and build something more meaningful. So i'm here :)
Welcome! it's an oddball little community here, lots of talk about Bitcoin and Red meat... but there are good folks and it's generally a good hang. But cultural dynamics exist here too, and I can sort of see the same thing where Devs wind up being community leaders, for better or worse. One thing I've been trying to harp on is better links to Mastodon. To me, Nostr's probably going to be too "weird and bitcoin-y" for the masses at the moment, but Mastodon is the place that can serve as a sort of refuge for both individuals and institutions that are getting "Enshittified" out of legacy platforms.
"When infrastructure providers bundle hosting with specific ideological positions or community identities, they create a new form of power over users." When I first heard Rudy talking about Blacksky on Rabble's podcast, my first reaction was "this should have been a Mastodon instance." This only solidifies that view. Like a lot of "VC Blockchains", e.g. Solano etc, they might be "decentralized" on paper but the cost of participating in the network is great. It seems like a similar thing with ATProto.
when I reviewed how the protocol that bluesky uses worked, I would say that it could offer decentralization, but the servers are like mini groups, where the network is only divided into parts and each person who hosts those servers is the one who has control of those groups, for example. so there the users are still subject to the rules that each server has, if you only change to another, you will not be free, it's just changing moderators.
As someone who spends time between Bluesky, Nostr, Farcaster. Interesting dynamics happening. Some folks were bummed Farcaster was β€œgiving up” on social. Then went ham on a user for asking about decentralization & censorship potentials Nostr accepts person in. Priving why decentralization good. Farcaster Build your own client of gtfo πŸ˜‚ Mastodon πŸ€·β€β™‚οΈ View quoted note β†’
I think the idea that the PDS servers "should" be neutral is unrealistic. If community servers are acting as de-facto infrastructure and you suppose that they "should" do something, well then, there's always an implicit enforcement mechanism standing behind the word "should". If you still need to invoke a centralized authority to make the players in a supposedly decentralized system work together, well, back to square one then. I will admit I myself have fallen into this kind of wishful "should" thinking before. And I think the reason why it's so easy to fall into is because the early internet had a lot of self-selected people that naturally aligned idea on certain values like curiosity, not assuming things about people, etc. But it's a different game now. I do think nostr is closer to a workable solution than bluesky, because nostr is fundamentally oriented toward redundancy/exit, not consistency/consensus. Notice that biological systems mirror this architecture.
Pushing for enforced common carriage (Title II style) on PDS operators misses the point of portability and competing providers. Those obligations only apply where users have no real alternative. If you can switch PDSes or self-host without losing anything, operators get reciprocal choice too. Take that away and you don't achieve neutrality; you just swap voluntary communities for coerced ones. Mutual and multiple selection is the real check on power here.
I've been working on a protocol design that could solve these issues. What I'm seeing in a lot of the critiques of social media infrastructure boils down to the power dynamics that emerge when infrastructure and community ideology are bundled, when the protocol layer and application layer are built together and are designed to depend on each other. Bluesky was designed FOR the communities on the "Blue" end of the spectrum (Western, liberal) as an alternative to centralized social media platforms that censor or allow infiltration by people with opposing ideologies, and the decisions on how the protocol layer works are colored by that initial intention. The goal was to create an alternative to Twitter, but not necessarily to address the underlying issue that plagues all of our social media options right now. Blacksky was explicit in trying to create a safe online space for the Black community, mitigating harms like racism. To do that, they built their own PDS, a custom Relay and custom feed generators. This tied the infrastructure design goals to the community the application was being designed for, meaning that the ideological purity is built into those deepest infrastructure layers. Nostr is much closer to the ideal, separating identity and message transmission away from any centralized authority. But the Achilles' heel is the same. Relays, like the PDS nodes of AT Protocol are volunteer silos with the node managers and hosts doing the work of deciding what to host, or giving up that kind of power and authority and letting it be a free for all without an economic incentive. Anyone can spin up a Nostr relay node and open it up to the world, or close it off to a specific set of users. I'm seeing a need to fill the incentive gap by creating a financial incentive at the storage layer that would guarantee neutrality, basically turning data storage into a utility service. This would allow community based applications to be built tapping into this storage layer and performing the ideological filtering and curation without the operator needing to feel the pressure to censor or gatekeep the storage infrastructure itself. They way this could happen is by using the metadata tags on the content itself, which would identify the creator, the topic, and whatever else the application being used is designed to add to the tag details. This would change the negative filter (I don't want XYZ) into a positive filter (I DO want ABC), without changing anything at the protocol layer itself.
I personally think that Bluesky's PDSs should have remained neutral infrastructure. Before, when they were like that, you could effectively build out your own platform using Bluesky's infrastructure for hosting, kinda bootstrapping the protocol and letting anyone build a platform with the rules of their choosing. Now that Bluesky moderates at the PDS layer AND the relay layer, and since they host the vast majority of AT Proto infrastructure, they've more or less locked the entirety of AT into Bluesky's community guidelines. (Who'd want to build a platform on AT with different guidelines, if no user of Bluesky - which is practically the entire network - could partake in said alternative AT platform?) As far as other platforms though, I'm more agnostic on neutral PDSs - if I understand the issue with that controversy it was with Blacksy. If an alternative platform moderates at the PDS layer that's not going to effectively apply the rules to all of AT like Bluesky can. Once a protocol is more widespread individual instances of it can have a diverse set of rules, but if one instance (e.g. Bluesky Corporate) can apply it's rules to the entire network before it expands, then the protocol could be captured and prevented from expanding beyond that initial niche base that captured it. My 2c anyway.