I don’t take issue with your points, but you’re conflating issues. I think it’s worth clarifying some terms up front.
Being utterly independent isn’t necessary for decentralization. Decentralization very specifically means there’s no single holder of the data; it does not have any implication for dependencies.
Lemmy is not decentralized; it’s federated. “Decentralized” and “federated” are not synonyms, and as long as you doing don’t run your own server, you’re effectively on a centralized platform. This is to your point about being “always dependent on something, somewhere in some way.” It’s true for Lemmy; it is not true for all systems, not unless you’re being pedantic, which wouldn’t be helpful: you being dependent on electricity from your electric company doesn’t mean an information network can’t be “truly” decentralized.
A distributed ledger can be truly decentralized. Blockchains aren’t always distributed ledgers, and not all distributed ledgers are blockchains, but whether or not a specific blockchain is resource intensive has no bearing on whether or not it’s centralized. This is the part I take issue with: it’s irrelevant to the decentralization discussion.
Bitcoin is decentralized: no single person or group of people control it, and there is no central server that serves as an authoritative source of information. If there were, it wouldn’t be nearly so ecologically expensive. Its very nature as something that exists on equally on every single full node is part of the cost. You can take out any node, or group of nodes, and as long as there’s one full node left in the world, bitcoin exists (you then have a consensus verification problem, but that’s a different issue).
But let’s look at a second, less controversial, example: git, or rather, git repositories. This is, again, fully decentralized, and depends on no single resource. Microsoft would like you to believe that github is the center of git, and indeed github is the main reason git is as popular as it is despite its many shortcomings, but many people don’t use github for their projects, and any full clone of any repository is a independent and fully decentralized copy, isolated and uncontrolled by anyone but the person on whose computer it exists. Everything else is just convention.
Nostr is yet another fully decentralized ecosystem. It is, unfortunately, colonized almost entirely by cryptobros, and that’s the majority of its content, but there’s nothing “blockchain” or crypto in the core design. Nodes are simple key/value stores, and when you publish something to Nostr you submit it to (usually) a half-dozen different nodes, and it propagates out from there to other nodes. If you run your own node, even if your node dies, you still have your account and can publish content to other nodes, because your identity - your private key - is stored on your computer. Or, if you’re smart, on your phone, and maybe your laptop too, with backups. Your identity need not even be centralized to one device. No single group can stop you from publishing - individual nodes can choose to reject your posts, and there are public block lists, but not every node uses those. It is truly decentralized.
I’m not familiar with Plebbit, but it seems to me they’re trying to establish a cryptographically verifiable distributed ledger - a distributed blockchain. There’s no proof-of-work in this, because the blocks are content, so the energy cost people associate with bitcoin is missing.
DHTs and distributed ledgers are notoriously difficult to design well, often suffering from syncing lags and block delivery failures. Jami is a good example of a project plagued by DHT sync issues. I’m not surprised they’re taking a long time to get stable, because this is a hard problem to solve - a deceptively simple problem to describe, but syncing hides issues like conflict resolution, updating published content, and all the administrative tools necessary in a world full of absolute shitheads who just want to cause chaos. It does look to me as it it would be fully decentralized, in a way Lemmy isn’t, if they can get it working reliably.
sugar_in_your_tea@sh.itjust.works 1 hour ago
That’s not exactly true.
That said, you do need some form of centralized service to connect peers, but you can federate those. It’s only job would be to connect peers, and a STUN server w/ TURN fallback is usually the approach here. These instances don’t need to store any data long term, they just need to connect peers, and the client is free to choose any instance they want, or host their own.
That’s how Tor works (entry nodes), and most decentralized systems use a similar system.
One of the best parts here is that offline often just works, and you can sneakernet around firewalls (e.g. if you visit China or something), and all you need to do is connect to a local relay to find local peers.
My understanding is it’s only used for name resolution, so the number of data points here should be in the thousands, not millions or billions, so the resource usage should be minimal.
Basically, the blockchain is functioning as DNS here.