Privacy is a reverse idea on the Fediverse. I know it’s a hot take, but by design the Fediverse is never going to be private and people should stop assuming it is.
When you send out a comment/like/post/whatever, you are literally broadcasting a message to any other instance listening. It essentially just says
{ messageId: 42, message: "This is some message", action: "comment" }and if you want to delete that message it’s essentially
{ messageId: 42, action: "delete" }While Lemmy and Mastodon respect that, anyone can build any fediverse app and simply choose not to use it. Anyone can build a search engine and can choose to respect the delete or not. Any instance could defederate from them if they don’t like that, or they may not care. The point however is that ActivityPub is designed this way, and there really isn’t a better way.
If your comment has been sent out to other instances - well then it’s there already. You can’t delete it without some form of just asking politely that they delete it. They have it already, it could be stored in their DB, duplicated in other DBs, aggregated and sent to AI, searchable, whatever. They have it. There is no concept of “delete” on the fediverse. It’s asking nicely for them to delete it.
The thing most people get wrong is privacy friendly =! private. If you say something publicly (on the internet) you can assume it will stay for ever, if not directly then via some sort of archive. The privacy part of Lemmy/Mastodon is them not collecting data on what you look at to sell it. If you want something private then don’t use Social Media, because what you say publicly will stay public.
deleted by creator
As you say though it’s only shared to any other instance listening. The point of consent-based federation is that you get to choose which instances do and don’t get to listen. So if your comment hasn’t been sent out out to other instances, they don’t have it.
So if your comment hasn’t been sent out out to other instances, they don’t have it.
What’s stopping malicious actors to create an account on the same instance as you and follow you (or your RSS feed) exclusively to pull your data?
Remember “information wants to be free”? That adage works both ways. If people want (or need) real privacy, they need to be equipped with tools that actually guarantee that their communication is only accessible to those intended to. The “ActivityPub” Fediverse is not it. They will be better off by using private Matrix (or XMPP rooms) with actual end-to-end encryption.
Agreed that people who need strong privacy should use something like Signal (or maybe Matrix or XMPP). And also agreed that RSS feeds are a privacy hole on most of the fediverse; Hometown and GoToSocial both disable them by default, Mastodon should do the same.
Nothing prevents malicious actors who want to make enough of an effort from creating accounts on instances (or for that matter Matrix chat rooms). But that’s not feasible for broad data harvesting by Meta.
Your whole wordlview is hinging on two conflicting realities:
- social networking is an inherently public activity, and this is the way that the majority of people want it to be.
- the only way to be free from surveillance capitalism is by having private communications, and while this is something that affects everyone, only a minority of people seem to be actively opposed to it.
The “consent-based” social media does not work well for a small business owner who wants to promote their place to their local community, or the artisan that wants to put up a gallery with their work online. They want to be found.
If you tell them that they have to choose between (a) a social network that makes it easier for them to reach their communities or (b) a niche network that is only used by a handful of people who keeps putting barriers for any kind of contact; which one do you think they will choose?
What your recent articles are trying to do is (basically) try to shove the idea that the majority should change their behavior and completely reject a public internet. You are basically saying that the “social” networks should be "anti-"social in nature. This is, quite honestly, borderline totalitarian.
But that’s not feasible for broad data harvesting by Meta.
Why? You keep writing about how evil Meta is and their infinite amount of resources. If you really believe that, why do you think they would stop at the mere wall of “federation consent”?
deleted by creator
I totally agree that there isn’t a lot of privacy on the fediverse today – in fact I even say that in the article and link off to recommendations for how to improve things. But also I think there’s a huge difference between the situation on the fediverse where there’s no privacy because developers haven’t prioritize it and with Meta, where their model is focused on exploiting data that they’ve acquired without consent and they’ve repeatedly broken privacy laws (although to be fair they break other laws too, not just privacy).
And it’s true, many people don’t care about privacy, and many more care some but it’s not important eough to them to make it their primary reason for choosing a social network. But a lot of people do care, at least to some extent, so the free fediverses will be a lot more appealing to them if they improve privacy. And even though I think privacy by itself won’t the major driver for most people who choose the free fediverses, improving privacy also works well with that I think will be the major drivers – like safety, pro-LGBTQIA2S+ focus, and (for people who want nothing to do with Meta) highlighting the core differences from Meta.
Circles’ approach is certainly interesting, I remember looking at it when they did their kickstarter. Did it go forward? It looks like their blog hasn’t been updated since 2021.
deleted by creator
Its documentation, for example, describes consent-based allow-list federation as “contrary to Mastodon’s mission.”
and I would agree with them. Consent based federation would fundamentally change the fediverse and create large tenants overnight. Small instances like mine would be at the mercy of large instances to be federated with them. It relies on people being kind and open, something we have already seen that some instance owners can be, others are not. I would even argue that that isn’t even federation anymore, it’s just slightly more open walled gardens
Yeah, as I say in the article Mastodon makes other decisions that are also hostile to the idea of consent, so I also agree that they see it as contrary to their mission. In terms of large tenants, though, Mastodon changed the defaults to put sign people on mastodon.social, which as a result now has 27% of the active Mastodon users, so I don’t think that’s the basis of their objection.
And no, consent-based federation doesn’t rely on people being kind and open. To the contrary, it assumes that a lot of people aren’t kind, and so the default should be that they can’t hassle you without permission. It’s certainly true that large instances might choose not to consent to federate with smaller instances (just as they can choose to block smaller instances today), but I don’t see how you can say that’s not even federation anymore. Open source projects approve PRs and often limit direct checkins to team members but that doesn’t mean they’re not open source.
I’m not saying that it’s not open source, I’m saying that I would argue it’s not federation anymore. Open source is irrelevant here, I’m not talking about the code.
I’m saying instances being “Closed to federation by default” and “whitelist only” is not true federation in my book.
I also am saying that instance owners are the ones who all of a sudden get a ton of power, specifically larger instance owners because they can decided arbitrarily not to federate with an instance they don’t deem worth federating with. The larger userbase aside, instance owners I believe can become power hungry and greedy and refuse to federate.
For example, even I, a teeny tiny instance owner, felt a pang of annoyance when someone created a duplicate community on their instance. It was fleeting and I told myself that that’s what the federation is, and that it’s okay, but not everyone will react that way. It’s inevitable that larger instances will say things like “Why should I federate with you, we have all of those communities over here”
My open source analogy wasn’t great, but the point I was trying to make is that even things we usually think of as open are compatible with consent. Similarly we’re used to thinking of federation as unconstrained (well except for Gab) (and everybody else who gets blocked) but that’s just the specific flavor of federation that’s been practiced on the fediverse so far -federation’s compatible with consent, at least in my books.
Power-hungry instance owners can already decide not to federate with other instances, arbitrarily or for any reason – counter.social’s an example. Consent-based federation just changes the default. It’s true that this changes the equation a bit; today there’s a small amount of effort required not to federate, a consent-based approach flips that and there’s a small amount of effort required to federate. At the end of the day, though, power-hungry instance owners are gonna do what power-hungry instance owners are gonna do; threads.net and mastodon.social are going to make their own decisions about federation policies no matter what the free fediverses decide.
It’s fine if single instances do consent-based federation that prioritize safety over openess, but why should it become the default for all instances? It will result in instance protectionism and an overall decline in discussion quality. Making it opt-in means people will connect less likely with folks from other instances, meaning people will mainly stay on their instances, meaning it supports tribalism in the Fediverse. More safety usually comes at a cost, too. In this case: less interaction with other instances.
But if you federate with instances that you trust good enough in the first place, constent-based federation is not necessary imo.
No wait, I was wrong Its not necessarily instance protectionism. For especially vulnarable groups consens-oriented federation might make sense.
The question is whether this is the desired state for all instances and I would disagree here. I think this falls under a bigger societal debate: should the fediverse become a place were all potentials of harm are completely erased? In other words: should the Fediverse become a safer space?
First of all, minorities should be protected as by the laws of many countries. However, what harm looks like beyond that should be dynamically defined in social debate. Now you want to skip that and erase all potential out of the stand.
This ignores that these societal norms change over time and that a certain risk is part of the human condition. There always needs to be a balance between freedom and protection for the whole society. But as said before, safer place are also needed, but they dont work as blueprint for the whole society.
Early christian groups can also considered safe places. You are aligned here with what to me are totalitarian argumentation patterns that thrive for a garden eden that will never exist.
That doesnt mean that we shouldnt thrive for certain ideals but not for things that cannot and shouldnt be expected of people, like giving up their free will for complete safety.
I agree that different instances will make different choices based on their priorities, but follow this through. Take trans people as an example of an especially vulnerable group that consent-oriented federation makes sense for – so trans people will be be less safe on instances that don’t take a consent-based approach. What instances do you think trans people will prefer to be on?
And there must be something I’m issing, because I don’t understand how you got from consent-based federation to “giving up free will”. Consent is literally about having the ability to choose, so exercising your free will.
@ada@lemmy.blahaj.zone
prioritize safety over openess
I don’t expect anything I post here to be private.
On Lemmy? Certainly not. But on other fediverse software, there are followers-only posts, direct messages, local-only posts … none of it’s encrypted, but still it’s not public.
this seems like nonsense. as if youre going to limit who can see your public posts… the fediverse is opt-out not opt-in. you opted-in when you signed up with an AP federating platform.
if you dont want to federate, dont use a federating platform. if you want privacy, dont use a platform designed for public distribution.
It’s not as much about seeing, it’s about using, and profiting from.
This post under CC-BY-NC-SA.
If something is on a public unencrypted website, it isn’t private.
Unfortunately certain people have chosen to mislead users about this.
You may as well post your ass on a billboard then complain that people look at it.
Fediverse software has followers-only posts, direct messages, local-only posts … Mobilizon and Streams even have private groups.
None of that is private. It’s all readable by anyone with an admin account.
As a general rule. If it’s not end to end encrypted, assume it’s public.deleted by creator
That’s missing the point though: if something isn’t completely private then it has the chance of going public. Too many services pretend to be more private than they really are by using terms like “private message” when all they’re really offering is a relatively small barrier to seeing your data, especially if anyone can set up their own instance.
deleted by creator
Tbh I’m struggling to imagine what this would look like in something like Lemmy. It seems to be describing an extreme form of setting your account to private, but this only really makes sense in a situation where you have followers who are friends and family. How would I decide who to “approve”?
Great point, I should be more explicit in the article. On Lemmy, it would look like a couple of things:
-
today, another instance’s request to federate is accepted unless it’s explicitly blocked. This means that bad actors can get away with stuff until they’re discovered and blocked (although it makes it easier for good actors to federate). Consent-based federation turns that around: a request to federate isn’t accepted unless it’s approved. One way an instance admin could decide whether or not to approve a request is to look at FediSeer to see what other instances are saying about the requestor.
-
at the individual level, it would mean that people would start out by participating in local communities (and maybe even just seeing posts from their instance, not sure about that), and could then choose to have their posts federated out
That sounds like it punishes small instances… a lot. What would starting an instance look like? Do you start with a huge list of servers to inspect and approve?
For new instances, the easiest thing is to start with the list of an instance that the kind of moderation you agree with. If I were starting up an instance in the Lemmy world, I might go with the current federation list of lemmy.blahaj.zone or beehaw.org (although others might make differnet choices), in the Mastodon world I might use awoo.space as a starting point.
There’s certainly a need for tools to make this more scalable. “Recommended lists” are a likely next step; there isn’t much software support for this yet, but it’s similar enough to blocklists that they’re also fairly straightforward; it would be up to the new instance admin to decide how many to inspect or whether just to trust the list. And tools are also needed to address the challenge in the other direction: how do existing instances decide whether or not to accept the request? Instance catalogs like fediseer can help. Another possibility that I mention and link to in the article is “letters of introduction”; federations of instances (which I’ll talk about in the next installment) are another.
-
