I briefly posted about this in this thread yesterday. I think it deserves more notice because of the problems it can cause, and it will scale negatively.
Currently, instance bans and community bans are treated as two separate things. When a user is banned from an instance, you’ll often see in the logs a bunch of community bans alongside it at once (at least from Lemmy communities). These are communities that user has posted on. An instance ban automatically applies hard-bans to communities they have interacted in from that instance. But the problem here is its only communities they’ve interacted in.
The instance ban itself is simply a rejection of federation. It doesn’t block users from posting in communities on that instance - only the community bans do that. It just means their posts won’t federate out. This means that an instance banned user can continue to be a nuisance in most communities (or all, if they are pre-emptively banned) on an instance locally - and the moderators of that community and instance won’t even know because they don’t view their community from there. With larger numbers of users would also mean larger amounts of trolls and incompatible users, which could greatly increase the chance of people simply vandalising communities and no-one even noticing.
Lemmy 1.0 promises to fix this apparently from their end, but I think at least for as Piefed is concerned we could get in on this first. We need a hard block on all Piefed accounts from being to interact on any community that is from an instance they are bannedfrom. We also need a way before that for Piefed based communities to automatically throw out all comments made even locally by instance banned accounts based from Lemmy.
To my understanding, only users from the banned user’s instance are seeing the banned users posts, as all other instances sync with the main instance. And if the main instance refuses federation, all other instances won’t get the banned user’s posts either.
Correct, but an instance banned user can still be a menace on their local instance in this context.
Isn’t that the whole point of sandboxing
I assume it is not the desired point that if I ban a user from piefed.social for posting gore (example) that they continue to post gore all over piefed.social communities from their local instance with those community moderators being completely unaware, meaning local users from there would see it.
Isn’t the idea to aggregate community names between instances. So piefed.flowers gets goretown.flowers. Piefed can defederate goretown, piefed.flowers can ban goretown.flowers, or any level can ban user Joe to hide it in their sandbox. But what piefed.flowers cannot do is stop Alice from seeing Joe’s gore in Flowers from goretown if both are from goretown.
Maybe I’m wrong, but that’s the consequence of flexibility and sandboxing.
If you are banned from a particular instance you can still see all posts and comments from that instance, but you shouldn’t be able to post on any community from that instance. That’s the problem.
We had this happen yesterday where a troll from a currently unmoderated instance signed up to it. They were banned from everywhere else, but they were still posting gore on communities they were banned from and local subscribers would still see it.
I think instance bans are way too complicated to understand. There’s the community’s home instance, the author is on an instance which might be different. And the reader maybe on yet another one. And we allow the moderators to be do it from anywhere. So there could be up to four instances involved. And anyone could be instance-banned on any of those. Go figure if a post could evade the community mods because it’s hidden from their view. Or visible to me as a user when it shouldn’t.
I feel as a user, I’d expect an instance ban to be the exact same thing as a ban in every local community. Show up a similar way in the modlog and everywhere. And of course their content on remote communities needs to be hidden as well. Unless I’m a mod, then I need to see it.
And I’d welcome if these things federated in a way. AFAIK community bans do federate. Instance bans don’t federate. So if for example someone gets banned because they molest other people, the outcome depends on who takes action in which role.
And there’s a dozen of instances I trust, and once their awesome admin team decides to ban someone, I regularly want that stuff gone on my instance as well. I think as of now it’s always up to a malicious user’s home instance to take action?! So spammers, trolls, bots… Might still be around on my instance until the specific admin team caters to them. I feel I could as well trust whatever decisions piefed.social (for example) came up with and reflect that on my website. That’s often way faster. Also faster than some (badly moderated) communities.
Hypothetical Auto-federated bans are another entirely, but this specifically is about plugging a hole with instance bans right now.
Sure. I just read the word “fundamental” in your title, and I know what what the motivation was in the linked post… So I felt invited to list a few other things i think are bad in the design of the ban system 😄
I mean combining/integrating instance-bans with community bans could make them federate like any other community ban. So on a first glance it seems to me it’ll just propagate anywhere and solve this issue entirely? Of course it doesn’t exclude any other (prior or more specific) patching.
If you get banned from an instance on lemmy now, you will automatically get community banned from any communities on that instance you have interacted in.
But only those. So you could still post to communities on that instance locally that you had not posted on prior to being instance banned.
Yeah, that should obviously apply to all communities. Plus be added to newly created (discovered) communities.
Makes sense to me.
How it works now, or the changes I propose?
Oh sorry :)
I mean I agree with your description of the problem and the changes you propose.
I’m coding it now, should be deployed today.
That’s great!
What about Lemmy users? I know if a banned lemmy.world user was to post on a piefed.social instance that they were banned from that it would be federation rejected, but what I propose here would go further and simply have it automatically purged at source too.
