I imagine users go poof. Are their profiles stored in other federated instances? Is there a way to recover them or “import from backup” onto another instance?

If they don’t have an e-mail I imagine you can’t even notify them or authenticate them elsewhere so this “import from backup” even if technically feasible (idk if it is) would be impossible in practice due to authentication issues.

And communities, can you even notify all your subscribers to move to the “backup community” on another instance? I saw yesterday that a Mastodon server host said “I’m deleting this instance in 2 days” or something like that and I started wondering how shit would go on Lemmy.

  • Skull giver@popplesburger.hilciferous.nl
    link
    fedilink
    English
    arrow-up
    6
    ·
    1 year ago

    There’s a cache on other instances that contains (some) posts and comments the instance received previously, but everything is bound to break when fetching new comments and posts fails. I wouldn’t rely on it.

    Lemmy/Mastodon/kbin/other ActivityPub servers are federated rather than fully distributed. That’s where the name Fediverse comes from (it’s not, like some people think, invented by some federal government). That means servers maintain accounts and data, but can interoperate between each other. This is different from, say, BitTorrent, which can operate almost entirely without servers (through the DHT) and still maintain data.

    Lemmy, kbin, Mastodon, and other Fediverse servers work very much in the same way email worked back in the day, when it comes to servers. You can exchange messages between Gmail and Outlook in the same way you can exchange posts between Lemmy and Mastodon, and back when message lists were a thing you could subscribe/post to lists on other servers as well.

    If Gmail dies without notice, you can still read the emails you’ve received for as long as they’re kept on your server (back in the old days of 1-10MB mailboxes that wasn’t very long!) but you can’t do much else with it.

    Lemmy’s local storage isn’t really meant for safeguarding data. It’ll probably keep messages from other servers for some time, but don’t expect to interact with them well, or to recover your account.

    Even in Mastodon, which does support account migration, toots disappear when the original server dies. Moving accounts works by redirecting other instances to the new location, with old toots left at the old place. If a server does after moving, the old toots disappear but the new toots stay up and your followers will automatically follow your new account. If the server dies without moving your account, everything is gone.

    There have been some improvements since the invention of email, though. For example, if an email domain goes down, anyone can register the expired domain and host a new server, pretending to be the people on the old server and receiving mail destined for their now dead email address.

    With (correctly implemented) ActivityPub clients (like Lemmy/kbin/Mastodon), this is impossible without a backup of the database, because every account has a secret key that’s used to verify data (posts, comments, follow requests). So, even if Lemmy.ml goes down and some asshole buys it to send spam from “trusted” accounts, they won’t be able to!

    I don’t think there are fully distributed social networks out there today. Even nostr, a Twitter alternative with a basis in blockchain, has the issue that if a “relay” you used to post content goes down, that post disappears. The problem is that all of that data needs to be kept somewhere, and there’s a lot of content to keep. Every message would need to be mirrored a few times because you don’t want to lose your messages when you drop your phone, but you probably also don’t want to store random people’s messages and have a server on your phone drain the battery serving them back to the network!

    Your best bet to protect your account is to self-host. That means setting up your own server that you manage, with your own backups and your own rules and accounts. This isn’t exactly a one click thing, you’ll need some Linux server knowledge, but it’s what me and many other techy Lemmy/kbin people do.

    • God@sh.itjust.worksOP
      link
      fedilink
      English
      arrow-up
      3
      ·
      edit-2
      1 year ago

      I don’t think there are fully distributed social networks out there today. Even nostr, a Twitter alternative with a basis in blockchain, has the issue that if a “relay” you used to post content goes down, that post disappears. The problem is that all of that data needs to be kept somewhere, and there’s a lot of content to keep.

      That’s exactly how Hive works by the way, each witness has a copy of the blockchain software, and to run a node and earn money from it, they must “sync” the blockchain and that means running all the blocks software to reproduce all blocks one by one. This process can take hours or days to finish. But afterwards, any node or group of nodes can die and as long as one witness node exists for every microservice (main chain, hive engine side chain for tokens, some metadata chain and there’s a few other things ppl have invented idk what for), all data is safe. So as you say, it’s expensive and a bit crazy to require everyone to host all the content available but when done it gives some safety to your data.

      Hive is not federated tho. It’s one thing, just hosted by many with 100% redundance.

      Your best bet to protect your account is to self-host.

      I was thinking of doing that, setting up my own server, but for some reason all the communities on the server I’m on (sh.itjust.works) are very small and the ones on lemmy.ml grow a lot, so it makes me think that it’s much harder for people to find communities hosted on smaller servers than on big ones? Maybe I’m misreading the reasons.> Your best bet to protect your account is to self-host.

      • Skull giver@popplesburger.hilciferous.nl
        link
        fedilink
        English
        arrow-up
        0
        ·
        1 year ago

        Hive is a pretty funny name because there are two Hive networks with social features that have absolutely nothing to do with each other. I’m going to assume you’re talking about the cryptocurrency based system.

        You can post and subscribe from your own Lemmy/Kbin perfectly fine, by the way. I’m on my own server and I’ve only created a single community (for a theoretical blog). What you do to follow a community on another service is search for !community@server.com (maybe wait a sec or search again if it doesn’t show up) and then you can join no problem.

    • marian@feddit.de
      link
      fedilink
      English
      arrow-up
      0
      ·
      1 year ago

      how much storage space would one need on average if you are a hobbyist self-hoster? If I make my own server and, say, it gets a bit popular with thousands or tens of thousands people, all posting text, images and even videos all day, I imagine I will soon run out of places where to put all my hard drives, not to mention the electricity bill…

      • Skull giver@popplesburger.hilciferous.nl
        link
        fedilink
        English
        arrow-up
        2
        ·
        1 year ago

        My instance has grown to a few hundred megabytes in size for the messages (almost all coming in from other servers) and then there are the files.

        That’s for about 422k entries in the database, so I’m pretty sure all the communities I’m following are the cause. At that point, I don’t think it’d matter if I hosted five people on my server or five thousand; their posts are probably going to end up on my server anyway.

        And besides, I don’t need a long term copy of all that data. I can probably purge data every week to keep the database to a few hundred megabytes or so.

        Going by this post a $12 VPS with 500GB of storage should do just fine for years if you don’t accept too many files. All of Wikipedia is about 20GB of text, but image files will probably be your biggest issue.

        There’s a risk of user rush, but Lemmy’s design allows for manual approval of all user accounts. If you don’t want to run the risk of exceeding your bounds (or dealing with moderation fallout) then only allowing you and your friends would be a perfect middle ground.