ICYMI: dustyweb.bsky.social last week published her reply to feedback on her previous blogpost that analyzed Bluesky's decentralization: https://dustycloud.org/blog/re-re-bluesky-decentralization/

My executive summary: 🧵 1/N

1. Running an indexer is expensive, saying "Wikipedia/Internet Archive can do it!" isn't a retort because these are big orgs with big budgets. There's an opportunity here for decentralized indexing, which is a Bitcoin-level hard-distributed-systems-design-problem nut to crack.
2. Bluesky would fail to meet it's UI/UX goals if replies to posts went missing. Missing replies are a feature on Mastodon. Not on Bluesky. This requires something more akin to global consensus, which FC Hubs do.
https://dustycloud.org/blog/re-re-bluesky-decentralization/
3. In the extreme limit, if everyone were to self-host their own PDS (personal data server) and everyone were to run their own Relay & Appview (to index everyone's PDS and run Bluesky independently), that would require everyone-to-everyone message passing...aka scales quadratically 👎.
2/

Software Engineer @ Cadence Design Systems. B4: Stanford EE / Illinois ECE

4. But most people are not going to host their own PDS or run Relays/AppViews for Bluesky. You can't have a 'big-world' social network without running 'big-world' infra. You could run a restricted Mastodon-esque cozy clique of PDSs, but anytime those PDS users interact with out-of-network context (comments, retweets, etc), you won't be able to render it properly.
3/

My take is that I mostly agree with the point raised above, but using cryptoeconomics to tackle some of these problems seems like kryptonite (cryptonite? ha!) to DWeb/IndieWeb people. Data publishing & long term storage is something we know how to solve using cryptoeconomics. Indexing, less so, but we have some first passes with marketplaces like The Graph & will continue to build with slashable guarantees on top of Eigenlayer, perhaps towards 'fully sharded indexer networks w/ slashing'. We've also already seen at least one team planning to run FC Hubs as an Eigenlayer AVS, which is also a great first-pass solution. Finally, low quality spam clogging your tubes & ddosing your endpoints has a hope & prayer of being nixed in the bud with some cryptoeconomic sybil resistance. More than half of my BSky follows in the past week have been spambots imitating large accounts that I follow. 4/

Anyway, I maintain that BSky is not decentralized: not because indexing is a hard problem per se. It's because providing zero cryptoeconomic incentives for anyone to run an indexer means unbounded databloat spam followed by browbeating to "just start the next Wikipedia bro" (read: nonprofit industrial complex megacorp in 501c3 sheep's clothing). It doesn't pass the smell test...we shouldn't need to run 16+ TB & growing archival nodes out of the kindness of our hearts to be a 'peer' in the network. Crazy idea: BSky PBC submits a slashable bond on Eigenlayer where they agree to be slashed if they don't properly index & make the firehose data available. In exchange, they receive API fees via direct onchain payments (no need for Stripe + API Keys involved). This would be an order-of-magnitude step change even though it seems like window dressing. Skin-in-the-game keeps incentives aligned. 5/5