Hi there,

I went through the documentation of GoToSocial and there are some pieces of information which confuse me. For example on the Deployment considerations, they state, that once you hosted a particular Fediverse service on your domain, you cannot switch to another technology. Further down in this article in the “Domain name” section it even gives me the impression that if you switch technologies on the same domain, this will in fact cause issues in the whole Fediverse.

Two questions came up when reading through this:

  • Is the ActivityPub protocol and the technologies that depend on it that fragile? Switching technologies on the same Domain would be something I would have just done without a further thought until I find the technology I want to use for years (and which I might still switch out to another one many years in the future).
  • It is not clear from the documentation if you can get around this by hosting the service I want to try under service1.example.com instead of example.com. The documentation states, that you can host your users under user@service1.example.com, but the API services still under example.com. This will not solve the root issue, right?

Getting a new domain for each Activitypub service I might try to implement and test / use does not really sound great to me. Maybe I just did not understand all of that properly and there is no issue?

  • squaresinger@lemmy.worldEnglish
    3·
    2 days ago

    It’s a worthy and huge endeavor.

    I think the only somewhat sustainable way to get around the moderation problem is to get rid of storing a copy of everything.

    That way you don’t have incriminating data on your server and then you just mark every external community as “out of bounds of my moderation, there be dragons, go at your own risk” and call it a day.

    If a Lemmy/ActivityPub alternative was designed from the ground up a decent option would be to limit federation to a single-signon and private messages. In that case when you visit a remote community, the client directly goes to that remote community to fetch data from there.

    Basically like a set of separate forums with a federated login.

    That would solve the “everything copies everything” issue and the “everyone has to moderate everything” issue as well. If someone posts illegal crap on a remote instance, that data stays on that remote instance and you aren’t responsible for them. And the users can themselves decide what communities on what instances fit to what they want to look at.

    That would mean that if an instance goes down their communities do as well, but that’s (at least to me) less of an issue than the current state. It’s not like these zombie communities work fine right now. With the source instance being down, federation is gone and thus posting on these instances means there’s only a fraction of the audience left.

    • sugar_in_your_tea@sh.itjust.worksEnglish
      1·
      1 day ago

      limit federation to a single-signon and private messages.

      But then you completely lose content when someone disables their account, and most people don’t want to host anyway.

      My approach is a P2P service that’s similar to that, but instead of storing your stuff, you store some amount of other peoples’ stuff, such that there are multiple copies of any piece of content distributed randomly across the globe.

      However, moderation gets tricky here. I think a transitive trust system can work well. Basically, Alice trusts Bob to some degree, Bob trusts Carol to some degree, so Alice trusts Carol to some lesser degree. If content falls below some trust level, Alice doesn’t see it or store it. Bob and Carol don’t even need to be people, there can be bots to detect things like CSAM and other illegal content.

      The net result is that everyone’s experience is tailored to them, which hopefully makes things like shilling, trolling, and astroturfing less prevalent for those who curate their trust network more effectively. And this curation doesn’t need to be manual, it can be automatic based on how you react to content. In other words, everyone is a moderator, and you trust people who moderate similarly to you.

      The intent here is to solve a bunch of different problems:

      • prevent content from disappearing
      • eliminate power hungry mods
      • reduce cost of operation - this only requires a handful of servers to help connect nodes
      • limit effectiveness of bad actors

      There are certainly issues, such as:

      • bad actors can use the network to share illegal content (everyone else just won’t see it)
      • hard to bootstrap - what does the new user experience look like?
      • what does an account mean without a central server? This is a similar problem as simplex has.
      • what’s the killer feature to attract regular people vs the criminals?

      And some interesting side effects:

      • since it’s P2P, you can sneaker net it behind firewalls (e.g. China) to facilitate global free speech
      • everyone’s device can add resources to the network, so the barrier to self-hosting is zero
      • interesting options for localizing data - communities for a region will be super responsive for people in that region because there are more interested nodes; could maybe find new communities by local popularity

      Given the downsides, I’m not completely convinced it’s worth it, hence the hesitation. But anything that requires users to have a publicly facing server is DOA, so this seems like the most approachable option.

        • sugar_in_your_tea@sh.itjust.worksEnglish
          1·
          23 hours ago

          Yeah, they certainly have some similarities and I’ll have to play with them to see if either will work for me. Nostr seems closer to Mastodon and not really what I’m after. Aether could work, though I’m not really a fan of ephemeral posts, and I’m worried about Democratic mods in an era of bots. But maybe both are close enough to what I want.