shazow
@shazow.eth
Latest musing on Bluesky "decentralization": A key flaw is that censorship resistance is not part of their definition/threat model for decentralization at the social network layer. This has two consequences: 1. Points of censorship create opportunities for the network to split into separate pieces with different privileges (thus non-uniform infrastructure required to support different privileges). 2. It's much more expensive to compete with Bluesky's AppView ($millions/month), because this is the point of censorship. If there was no point of censorship, then more infrastructure could be a homogenous collective good that can be shared rather than substituted. Also relevant thought experiment: https://bsky.app/profile/shazow.net/post/3lbrtli7be224
2 replies
3 recasts
16 reactions
shazow
@shazow.eth
The less censorship resistance we have, the harder it is to be credibly neutral. More and more decisions fall on "the social layer" to decide what is allowed/disallowed, which will inevitably be arbitrary (like getting suspended for liking an unrelated post of an account that got banned: https://bsky.app/profile/haosov.bsky.social/post/3lbqa7ip2gs2c) This creates opportunities for negative-sum arbitrage as these "privilege networks splits" occur. Imagine X implements atproto and "invades" Bluesky. X doesn't have to be credibly neutral, so they can just filter out all the off-site content, but Bluesky wants to pretend to be credibly neutral so they would have to justify filtering the very demographic they're counterpositioned against. I don't think there's a win-win in these situations. Either neutrality erodes from the asymmetry, or censorship resistance has to be fixed (restoring symmetry).
1 reply
0 recast
2 reactions
Gabriel Ayuso
@gabrielayuso.eth
tbh there isn't much point in exploring Bluesky's decentralization, it is not, and I doubt it'll ever be. They've settled on handling Twitter scale traffic and provide some form of exit with PDSs but even that and their identity system isn't great. They use key signing but most PDSs hold the private key option for OAuth instead. There are some good ideas in the design. I especially like the lexicon and personal data store approach. It's something I was thinking about a while back. Beside that I don't think there is much.
1 reply
0 recast
2 reactions
