1y ago

Welp that answers a lot of why all .ml are down

https://very.bignutty.xyz/notes/9hf13it1ced3b2za

Furry Technologists @pawb.social

x-post

40 20

You're viewing a single thread.

523 comments

this is why instances should be abstracted away as underlying infrastructure and the users don't have to think about "instances". accounts and communities are replicated across servers.
- This is not the solution! Being able to pick a server to trust your data and content moderation with is a feature, not a bug.
  What we do have to do is make this feature more resilient and easier to use. Like adding the ability to easily transfer accounts and communities between instances, or even change the domain name of an entire instance.
  
  no, you're misunderstanding. that shouldn't be how it works. there shouldn't be any difference between the software on each instance such that it make your data insecure. this is how bitcoin works. this is why anyone can spin up a bitcoin instance and have it start contributing to the bitcoin blockchain and you as a user don't have to "trust" that particular node. trust is built into the distributed software architecture. you don't "choose" a set of bitcoin nodes. you don't "choose" your CDN or DNS servers.
  
  Cryptocurrencies and social platforms are completely different beasts. In crypto I want no moderation/censorship, I want anonymity, and there is a payout system so nodes can compete for something. This is all different when building a social network, so you can't just use the same architecture. Building social structures and trust is desirable in a public forum, not something you want to get rid of.
  
  This is all different when building a social network
  wait you want censorship in a social network? also, the architecture i'm describing does not do away with moderation and social structure. what about it makes you think that to be the case?
  
  Of course! Moderation is censorship. There is certain content I don't want to see, and I don't want to have to filter it myself so I join a community of seemingly likeminded people who censor content based on rules I generally agree with. They ban users who break the rules, keep spambots out, block malicious instances and so on, and if they are doing their job right then it builds trust and attracts more people.
  what about it makes you think that to be the case?
  Because you want to strip all that out and abstract it away. Who do you think would do the moderating and spam blocking? Who aggregates posts from all over the world and presents a sorted list to a user on their smartphone? It would be the wild west with users having to do everything themselves. I know it's tempting to think about building a Fediverse without instances, but afaik you need these social structures for the system to work.
  Crypto for example only works because you can define the rules mathematically beforehand, and then hand out money for computers to check them. That's just not possible with a public forum, at least not yet imo.
  
  you want to strip all that out
  i do not want to strip out the functionality of communities having mods that moderate the discourse and ban malicious users etc. it sounds like you misunderstood what i was proposing.
  
  What I'm saying is that you cannot do those features with what you're proposing, regardless of what you might want to do.
  
  ok. so you are misunderstanding what i am proposing then.
  i can explain in more detail any part of the design if you wish.
  
  Sure 😁
  I think the one thing we could adopt from crypto is having public keys as user IDs, instead of tying it to an instance. But that would require users to handle their own keys, and people are just reeeaaally bad at that!
  
  you connect to some lemmy instance on your web browser
  the client application (lemmy web app) authenticates your login credentials by first checking its own user database, if it doesn't find you (which it should because by default you'd be connecting to an instance that you've already used, and if done through a mobile app for example it would automatically find the best instance to use by lowest latency), it send out a message to the nodes(instances) that it knows about, searching for your user, recursively, when found, sent back and stored in each node that was part of the searching. (there'd be some threshold of tree depth so the unsuccessful branches don't keep going forever, and some other algorithmic details to prevent redundant network activity)
  you navigate to your subscribed communities feed, lemmy shows you the posts that are already on the node that you are directly connected to, then asynchronously sends out a request to the surrounding nodes to pull more posts from those communities, recursively reaching out to adjacent nodes, again avoiding repeatedly hitting the same node via algorithmic details which we can discuss further if you wish, sending back the info up the tree to your primary node. now a bunch of servers have duplicated community data, like a distributed storage system, but you, the user, don't know about all that stuff that just happened behind the scenes. your GUI is updated accordingly
  now you can interact with these posts, make new posts, and each interaction will be sent out to all the relevant nodes in a reverse process.
  another user on the other end can visit some community that you just posted to, and a request will again be propagated through the network, but starting from his node, and eventually reaching some node that has your new post.
  the advantages of this:
  if a node goes down, not all of the community and user data is lost, because its neighbor nodes have replicated the data
  if i am hosting a node, and have limited bandwidth and storage, i can specify limits so that my network is not unintentionally DoSed. so this implies that when the prior-described processes are occurring, some instances will not store the data they are pushing through, which is fine, and one of the intended features of this distributed architecture
  similar to previous point, each instance can have a whitelist or blacklist of communities (for either storage and/or data passing), defined by the admin, if he/she wishes to tailor the content for example to keep it related to content they are interested in rather than being forced to serve everyone on the network. it's like if someone wants to help a little bit but they don't have all the bandwidth and storage in the world, they can, instead of having to handle traffic for a bunch of irrelevant-to-them communities.
- There are a few technical problems with that. First of all, the cost of each instance would become quickly unbearable since everyone has all the duplicated data.
  Second problem, a malign entity could just come, create its own instance, spam everything and everyone with ads or whatever and suddenly every instance is full of that stuff. Also, how do you handle defederating in that case?
  What has been proposed before instead was to make some kind of mega communities that gather all posts from communities with the same name across instances
  
  everyone has all the duplicated data.
  everyone does not have all the duplicated data. they only have the data that they need -- the data requested by a user who happens to be using some instance.
  handling defederating is a good point. there could be malicious nodes that would be damaging to the network. i suppose there could be a community-mainted ledger of known malicious nodes (similar to minecraft usernames of known hackers), and the admins of the servers would maintain a blacklist. (obviously you configure that your instance's blacklist would be automatically synced with this ledger)
  the mega community idea could be good. where is this being discussed?
- I don't want to share an instance with the nutballs on the tankie instance or the nutballs on the fascist instance.
  
  to expound:
  the tankie instance or the nutballs on the fascist instance
  here you reveal a conceptual misunderstanding, or rather, a part of the lemmy architecture which i disagree with. there shouldn't be a concept of a "interest X instance" etc. it should be similar to a distributed storage model. so the concept of a community is not per-instance, it's just an abstract thing that exists in conceptual space.
  
  I'm aware of how you think it should work (Usenet, basically) but how does moderation actually work on Lemmy? Can someone be banned from a sublemmy on one instance and not banned from it on another?
  
  you already share water with them though. how is this any different? more seriously though, you already share internet infrastructure with them. the packets you just sent to make that comment could have been sandwiched between a "tankie" and a "fascist nutball". that's just the way it is man, there have always been crazy humans.
- This was my thought as well. Before learning more about the fediverse, I thought things are distributed and are replicated across servers (much like how distributed storage and computing works). But apparently they're not. You still have to choose which instance you want to use as your "home", and your data and your contents stays in your home. Others get to look at your profile and contents thanks to ActivityPub.
  I understand the needs for multiple instances (i.e., preferences for moderating concents, governance, etc.) But shouldn't the users and the user generated contents (arguably fediverse's valuable resources) should be safe-guarded by having redundancies in place across multiple instances?
  Has there any work or effort on this?
  
  I thought things are distributed and are replicated across servers (much like how distributed storage and computing works)
  yes, exactly! when you use the internet, you don't manually choose which ISPs to route through. you can pick which DNS servers to use but you don't have to. when you use youtube, netflix, or facebook, you don't choose which CDNs to use.

523 comments