Private and Public Mastodon
tbray.org
tbray.org
It's going to be interesting watching the norms of the fediverse collide with the expectations of ordinary software users. Mastodon and ActivityPub have a lot of promise as an alternative to centralized social networks, and there are good reasons to use them even if you don't buy into the ethos of the fediverse. If Mastodon is going anywhere, those users will soon drastically outnumber the keepers of the fediverse norms.
What is one of them?
It also makes sense that you can stand up your own Mastodon, in the same sense as you could stand up WordPress back in the day. Most people won't, but the fact that you can, and that the ability to do so is baked into protocol, keeps development interesting. It also makes it more interesting for third party clients, which Twitter has spent the last decade trying to eradicate. So that's a second (and probably third) reason.
Perhaps ironically, these are also reasons that I think the norms of the "fediverse" aren't going to matter in the long term. I think the software approach is important, and the notion of a single "decentralized social network" is not.
But regarding Mastodon as a potential alternative to the the current centralized social networks-- fair enough. I even saw recently that Ted Unangst even made a stripped-down Mastodon server in Go which I'd like to check out when I get a chance.
* You can write full posts in Mastodon
* You can own your own data, run your own service, or select from a variety of hosted options
* The ecosystem encourages good third-party clients and tooling
The tooling isn't there, but then, I don't think anybody expected the events of the last few months, so I'm inclined to give it a quarter or two. As a casual user, there are ways in which Mastodon is superior to Twitter already, and most of my problems are with the clients; Tapbots can't keep up with the beta demand for Ivory, and there's 2-3 other promising clients, so I'm optimistic.
I wrote Mastodon off for years, but: it's pretty good? I miss the blogosphere, and Mastodon is a step back towards it.
Do you see it as beneficial to have a series of large groups of balkanized servers optionally interconnected rather than a general public network everyone contributes to. I hate the label of "public square", because Twitter should already be opt-in in terms of which content you see based on who you follow, even though it's one big network (ignoring the @jack era AI-feeds which fought this idea).
A key difference in Twitter vs Mastodon UX is that users would have to make different accounts for each "Balkan country" server they join. If one server blacklists or is blacklisted they won't be able to participate in others. And this has implications well beyond US polarized politics once it reaches global scale.
I'm on a server (infosec.exchange) that I like just fine for now, but I'd expect to be publishing on my own server within a month or so. At that point, what's the "bubble"? It seems like exactly the same situation as RSS, except I'm not depending on Google to keep Reader going.
I'm not bullish on "the fediverse" as a federation of communities with some global norms (like religiously posting alt tags on images) and lots of local norms (like never posting AI art, or things like that). But having actually used it now, I think the underlying technology is a lot smarter than I gave it credit for, and I think it stands a decent chance of being important.
I personally would prefer going back to the "old" internet where we had niche forums. Mastodon provides that while letting forums connect with a wider set of communities.
I guess my only critique is that it's not often pitched this way. But it will become apparent to most people later on.
I think that if Mastodon sustains the growth they've seen over the last few months, then a year from now, it's going to be less common for people to primarily publish on communities like journa.host and FediScience, and more common for them to publish from huge Blogger-style community services --- with very few weird norms, much more comparable to 2021-era Twitter than to Mastodon.social --- or their "own" instances, just like everyone does with blogs.
That's interesting. Like I said, this bit will become more apparent to users as it grows, since in the current form that's all it really is (a few sets of big networks). I think you just sold me on the general idea of Mastodon.
Back to your earlier point, it might really come down to the UX issues around communicating on Mastodon itself, if it's going to pull this off. But that doesn't mean there won't be another service that can in the future. Otherwise that's the big thing that scared me away, since I'm a hyper-critical of UIs.
There are hosting providers doing that work for you, though several of them are overwhelmed too at the moment. There'll be more.
I run my own, and it's not hard but it's certainly still too much hassle for casual users, so I think there's definitely space for more hosting services.
> I'm not bullish on "the fediverse" as a federation of communities with some global norms
Yeah, a lot of these norms will collapse in the face of a broader user base that want Twitter 2.0, and don't care about a lot of the ideological stuff going on with Mastodon and other current Fediverse software.
> But having actually used it now, I think the underlying technology is a lot smarter than I gave it credit for, and I think it stands a decent chance of being important.
Treating ActivityPub w/WebFinger as a globallly addressable signed pubsub queuing system in itself is very interesting. You can build so much more on top of that than "just" "Twitter 2.0" or Mastodon. Of course there are some examples of that, like Pixelfed, but I think we're just scratching the surface of reimagining services on top of a layer like that.
Nobody wants to get shouted at by people preaching stuff they don't like or feel harassed by all day long, so we seek together in social groups that by and large are united at least along some axes of interests, and avoid too much talk of the topics we know are controversial within those the groups.
Some of us then sometimes seek broader forums to try to spread our groups ideas. But a lot of people actively avoid those kind of situations. Many others use different identities in different groups because they don't mesh well.
"Balkanized" servers are only balkanized to the extent people wants them to be (yes, admins can ride roughshod over users, but users can vote with their feed and migrate their accounts and follows elsewhere). Most people want to be widely federated. Some want forums where they can say things nobody else wants to listen to or wants to self-isolate (e.g. Truth Social, Counter Social and Gab all run - or at least ran, not kept up - Mastodon, but either don't federate at all and/or are widely blocked)
> A key difference in Twitter vs Mastodon UX is that users would have to make different accounts for each "Balkan country" server they join. If one server blacklists or is blacklisted they won't be able to participate in others. And this has implications well beyond US polarized politics once it reaches global scale.
If you want to participate in things so offensive to so many people that they're mutually blocking each other, then yes you either have to do that or find an instance that haven't defederated either. Or host/rent your own and not block anything other than what you, yourself choose.
With Mastodon you have that choice. With Twitter you risk getting outright banned site-wide.
This is true when you think it's about HN users as the prototypical user. But I'm just imagining having to explain to my mom why she can't participate in [X, Y, Z] set of Mastodon communities, doesn't she already have a Mastodon account? Why does she need accounts for each? Why are her old "tweets", followers, DMs, etc not available on the new server?
This is a much larger barrier than advanced users can appreciate... maybe not right now but it will increasingly be in the future.
Ultimately it's very possible groups of interconnected communities is a superior concept to centralized social networks. But this should be more of an obvious part of the UX w/ Mastodon. It's fine if it's properly communicated AND if porting your old tweets/accounts/followers/etc over to new servers is easy... but even then it's an adoption hazard (people avoid complexity they don't understand).
Currently it's communicated as a Twitter of FB competitor, which may very likely come back to bite them (not to mention the UI sucking).
"X and Y don't like each other, so they refuse to talk to each other. Think of it like how you can't talk to people on Facebook from Twitter". People siloing themselves is not a new concept for people.
Ironically, the new thing with Mastodon is that you don't need accounts for each of X, Y, Z to talk to people there even if X, Y and Z defederate from each other, as long as you can find "neutral territory". Only for a very small set of instances is this a challenge.
Being used to silos seems to also be behind a lot of the confusion over having to pick a server - people end up fretting over the choice when it matters a lot less than they think. I wish joinmastodon defaulted to just suggesting a random "generic" instance, and hid the rest behind an "advanced users" button.
> Why are her old "tweets", followers, DMs, etc not available on the new server?
Her follows and followers are available on the new server if she moves the account. When I moved to my new instance, most followers had auto-migrated within minutes. It can definitely be made simpler, but given the number of non-technical users I've seen migrate, it's not that big a barrier.
That the old posts are not I agree is an issue - it is likely to get solved, as the objections to migrating those with the rest are minor and stupid (you can migrate those too, but it requires a friendly admin willing to do a data import, or running your own instance, so for all practical purposes most users can't)
> (not to mention the UI sucking).
There are multiple UIs competing to improve this, before even considering the Mastodon alternatives that all can talk to Mastodon (e.g. Misskey, Pleroma), so this is improving rapidly. Personally I use the advanced interface and I think all it takes is some custom css and it'd be superior to Twitter in most ways (the advanced interface is somewhat like Tweetdeck; I run it with custom CSS that makes column widths configurable). Because of the total open nature of the API, there's far more room to get something good here than there is for Twitter, which is fairly awful.
Sure, but this is functionally no different from the days of forums when you'd have an account on many different forums to discuss the topics of interest to those forums. Just like forums, they may all be running the same two or three codebases, but the communities are different. The cool new thing federation adds on top of this is that communities can organically merge and split while retaining a basic level of cohesion.
And since they're all running the same protocol, it's straightforward for clients like Fedilab to make multi-account use easy.
So, I'd kind of question whether this is a real thing in any important way. In the early days of the Twitter exodus it _did_ look like it was going that way a bit, but in practice most instances quickly sorted out moderation to a reasonable degree. There certainly are largely defederated instances, but they're probably not of much interest to most users.
I would have some concerns about what happens if Twitter goes down for a protracted period and there's a much larger less self-selecting exodus, but for the moment at least fediverse balkanisation feels more like a theoretical problem than a real one which is likely to impact significant numbers of users.
One other potential looming challenge, I suppose, is Tumblr. If that does add ActivityPub support, it'll be, in effect, a very big, extremely poorly moderated, yet culturally very important instance, and how instances treat it could provoke a split.
I've been running my own Pleroma instance for a month or so and I'm firmly of the opinion that a single network is actually a Bad Idea. The fact that ActivityPub is federated means that it's also prone to abuse from other nodes similar to email spam, which is just about as simple to execute in the fediverse right now as rolling a new domain name. Small-time casual admins like myself won't be able to keep up with a global abuse firehose which hasn't really happened yet to my knowledge but will sooner or later.
The solution in my mind (and what I've already done) is to use a node whitelist instead of relying on blacklisting, which means smaller networks based foremost on interpersonal trust. Which means a constellation of much smaller but also stronger social networks that reflect actual communities as opposed to just the internet at large. This is super easy to do with ActivityPub servers and my prediction is this is the way forward although I expect people will also try to write clever abuse filters and fall into that perpetual cat-and-mouse game since that's what happened with email.
If you don't want your stuff archived and indexed by strangers you don't trust, then don't post your stuff in public to a federated network operated by strangers you don't trust.
On the other hand it's not really a technical problem, it's a social problem, contrary to everyone who will try to solve E2E federated group encryption in this thread.
It's a social problem because if I fork the project to enable search and quote-posts, any instances running that fork will get defederated by the community that doesn't want search and quote-posts.
Maybe the Eternal September of new people will want these features, and a fork will become more popular than the original branch. The people will decide what they want and vote with their instances.
As for me, I want self-search, so I built my own full-text search service that indexes everything I post and everything I tag as interesting. I've opened it up to a few friends, but we won't publicize it since it would be a pretty unpopular thing at the moment.
And even when it is installed, it doesn't do what I want.
In addition to my own posts, I want to index the public posts of those I follow, and I want to organize by topics that aren't found in the text of the post, and I want to "follow" search terms similar to how the latest Masto allows following hashtags.
None of this is available with Elasticsearch even if it was enabled on my instance.
I'd disagree. It's fundamentally a [techno]logical problem - a flawed design that only "works" because it's not happening at scale. The pen is mightier than a sword, and ink lasts longer than memory. Those millennia-old maxims still stand very true.
It guaranteed to get only "worse" as the system scales up and gains more "bad" actors that don't conform to the behavior. And it never will be solved socially (only mended) just like email spam was never socially solved, despite giant collective efforts like RBLs, community spam filters and IP reputations. Only shoved under the rug or plastered over, to be repeatedly revealed when a next drama happens.
I'd be happy to be wrong (if you know that I am - please give me an example), but I think that social and legal methods are never the solution, they don't ever fix anything (unless we're talking about humanity radically changing - then, maybe). They provide certain safeguards against honest people not accidentally screwing up, they have chilling effects, and they provide some recourse after the trust is broken - but they never fix the core issue.
Sometimes there is no way to change the design. Like - as a possibly silly example - we can't fix theft or murder, so those are social problems because we don't have technologies (such as achieving post-scarcity or immortality) to fix them. But the information security here is essentially math problem and it can be fixed.
Of course, whenever it will be is a matter of cost. Chances are, no one would ever do it, because it'll be too complicated (=costly) and the social approach would be good enough. Like how the world can't even switch to the IPv6. But, again, something like that doesn't really solve anything for real, merely lowers the frequency of occurrences to a tolerable level.
Just to be clear we're talking about indexing + searching what people post on public internet feeds?
IMO there are only social solutions to this problem. Any legal one would probably kill any federated service that can barely compete with centralized ones as it is... or (more likely) not be enforceable in the first place.
There's two social solutions:
1) don't post things you don't want recorded forever on public networks
OR
2) stop treating every mistake humans makes in their personal history as deal breaker for life
The first is a matter of personal responsibility, which is not a very popular concept in certain communities popular online. The other is more rational and practical but is still a very serious problem in western culture. Countless people have been 'cancelled' (or attempts have been made) based on 10yr old internet posts the posters themselves denounce.
If you want technical solutions use Signal, not a federated Twitter competitor.
Yes. I've only brought legal here, as there is some movement towards using legal frameworks for privacy ("right to be forgotten" etc.), as I totally misunderstood what kind of social solutions you propose.
> There's two social solutions: [...]
D'oh! Now I see. And I'm honestly sorry, as I completely misunderstood what you meant saying it's a "social problem".
I wholeheartedly agree now - either of those would be an ultimate solution. Not just a solution - I also do believe that this is how things should be.
It can be argued how realistic it is - I mean, it's quite a giant change to how society sees things. But I hope - at the grand timescale - it's well achievable. Many societies had drastically changed during the 20th century, so who knows where we'll end up...
Quote-post can be a problem because it encourages bad behavior among people who aren't actively malicious. If the most popular Fediverse software doesn't support it, there won't be a widespread quote-poste -> pile-on phenomenon even if a few servers support it without getting widely blocked.
Search becomes a problem when a group of bad actors use it to harass and dox people. It doesn't require widespread adoption, just a small number of bad actors with access to an index. From what I've read about previous organized far-right harassment campaigns, I find it likely that bad actors are already indexing Fediverse posts, but not publicly talking about it. Excluding search features from the most popular software will not help with this. Blocking servers that admit they're indexing posts for search will not help with this.
Anything I publish on any social network I'd be also comfortable seeing posted on every wall in foot-tall letters, with my real name attached to it. Were I not be comfortable with that, I won't post it.
I consider all "private" or "limited" posting provisions on all social networks as paper walls at best, one small devops or SWE mistake from being torn and becoming visible to public. Maybe I'm overshooting a bit, but my past experience shows that it's not much of an exaggeration. So I never use any "private" features and post everything I ever post as openly as possible.
Social media is a colander. Expect anything and everything to leak. Embrace being public.
If you want something with more real privacy, as in not likely to leak if all parties involved take reasonable care, use something that was built with privacy in mind, like, IDK, Signal, or maybe a private Telegram chat, or a private Mastodon instance only accessible through a VPN which you personally and competently operate. Of course it's unlikely to hide you from state-level actors, but at least it's more likely to protect you from simple operator blunders.
You misunderstand. It's layered defense / minimizing exposure. Nobody is suggesting it can't be done, but it not going to be normalized in the current Fediverse, and most potential attackers do not care enough to build throw own search engine.
The main reason is quite clearly stated in the article, and I'm not sure if you didn't see it or just don't care because you're not affected.
> And, if you’re vulnerable, attack you, shame you, doxx you, SWAT you, try to kill you.
A lot of people that are being targeted by the likes of KF or Libs of Tik Tok find nice comfortable corners on the Fedivserse because they're not immediately discoverable by someone typing their name into Google.
For the defense in depth / layered defense to work, the initial assumptions should be different. They should be like Signal's, not like Mastodon's. The whole approach should be paranoid about not letting things seep through by any chance at all. I don't think that such an approach would be seen as productive by most users and developers. So, I view various access control settings in networks like Mastodon as a convenience filter, not as a security mechanism.
> not immediately discoverable
This is actually both important and achievable! If you never post your public identity, and things which easily link you to your public identity, it's pretty hard to doxx you and harass you IRL, even if everything you publish were publicly accessible. If you don't advertise your presence where everybody is looking, you have a good chance of not being harassed online, because the harassers won't try hard enough to infiltrate remote corners of the internet.
A reliable pseudonymous identity is a great mechanism, if you maintain reasonable opsec. I just won't expect a great level of insulation of your content from fellow pseudonymous users, and occasionally even from general public.
The entire model doesn't even rely on being entirely undiscoverable - the point is that you don't need to step back from public life because a bunch of weirdos like stalking you, it can be as easy as having a publicly visible profile with a lock on it. This prevents following, but unlike Twitter you're not entirely excluded from making public posts that anyone can interact with. You can just do a baseline of filtering when accepting followers and deciding which posts to make public or private. You'd be surprised how many of these stalkers and harassers can't make a profile that passes a basic "vibe check" - they either try too hard at putting together a profile or they don't bother at all.
Fedisearch and related projects are frowned upon because they would make it much more trivial to find people who are not yet targets based on a soundbite and to go through someone's "permanent record" (most implementations do not backfill posts made before the user was initially followed from an account on the same instance).
Mastodon's instances admins have blanket read access to every post and message, even the "private ones", yet they are not under scrutiny of RGPD regulators like for example Twitter. Are you sure those admins take good security measures, like storing backups encrypted, keeping their tech stack minimal and up to date, using safe configs? Do you trust them for not reading and sharing your conversations?
Regarding the Fediverse search, it's even worse as you're supposed to trust the rest of the world for not abusing it and indexing the entire network. Well, that's not how things works. With this kind of large scale systems you should know that if there is the possibility of an event, that event will happen very soon if not already. Either you impose restrictions or you accept the tradeoffs.
I think it's unlikely that you can prevent ML model building with a carefully designed license. The most common legal position (though not something that has been tested in court yet) is that training models is sufficiently transformative to count as fair use, and does not require any sort of license to the data.
You can see this in all the state of the art tools that are trained on all the publicly available data that they can scrape, without regard for license: translation (text), GPT-3 (text), Stable Diffusion etc (images), Co-Pilot (code).
For preventing trolling and harassment a licensing approach is an even worse fit, since those are not people who care about respecting licenses.
https://www.twobirds.com/en/insights/2021/singapore/coming-u...
For the case of the EU, based on the way the GDPR and related digital laws are drafted very much in a "spirit of the law" and with individual rights and agency trumping corporate interests, it seems fairly likely it would not just allow coopting personal social media content over the explicit wishes of the creators, regardless of whether any access was deemed legal. For the moment, I think the UK digital laws are still basically just copies of the EU ones as well (with some search and replace) but I guess they'll drift apart over time if there's no de-brexit.
It's also worth noting, especially given how it's written about in the linked post, that these are generally assuming that these uses do not destroy the market for the original works, and they come from the reality before all the art generation. The specific implementations everyone is talking about, especially in the art space, have now pretty definitively proven that we're in a new reality now.
It's also possible to send messages to part of the followers as well, some instances like qoto.org support circles. You make circles from your followers and post to just them.
I wouldn't call these privacy features, but ability for sender to choose what it wants to say for certain group of followers.
This is pub/sub, it's not pull based, so every time you release something it is pushed to the subscribers, your followers servers.
I would like to have a bit of pull based things as well, but ActivityPub is not built for it.
Even though it's pushed to a server, doesn't mean it goes to everyone in that server.
Right, the issue being raised is that there's nothing preventing it from going to everyone on the server except the server being nice about it.
It's in the specs what servers should do if it receives a message. If someone sends a message addressed to just these people the server then delivers it to just these people?
Of course if you send something from foo@gmail.com to bar@yahoo.com the server at yahoo.com decides who gets it, hopefully just bar.
In fact email analogy in my mind is most apt and easiest to explain, including the fact that it's not E2E encrypted. It's totally up to servers who gets your messages when you send them.
Stuff you post will only be visible to the set of users you want it to be visible to. Writing software to present those posts in a suitable UI is left as an exercise to the reader. And it's a bit late to start now, but I really wish we'd been able to start our fediverse journey with something that's at least supposed to be secure, then opened up the stuff that's intended be open.
I don't want to start discussing what Fediverse should be, but I have had this discussion in past, few points:
- ActivityPub has two signatures: HTTP signatures, which are ephemeral, and more persistent JSON-LD object signatures. They are contentious. Some servers disable JSON-LD object signatures, because they allow to proof cryptographically that someone posted something, thus deleted posts become "liability".
- E2E encryption will have similar ramifications, because not all servers agree that crypto be used.
Currently only way to do "privacy" is to deliver messages to just the people you want, and hope servers won't send them to third parties, like in email. That's what Mastodon does.
E-mail has decent brand separation between the protocol and providers. People don’t talk about joining .social, they talk about Mastodon.
Sure. But users are cognizant they're using Gmail. And when something goes wrong, they're halfway decent at attributing it to their or the recipient's provider. Mastodon servers are eclipsed by Mastodon per se. It's closer to AOL than e-mail.
Follower-only posts are not about hiding something from prying eyes, but about removing noise and clutter from those who don't care about certain topics.
Anyone thinking their post is only seen by their followers when they have federation turned on is grossly misinformed.
I think this is simply a social problem — when sending a post to your followers, you have to trust that they do not share your post. The same applies in private messaging. You have to trust the recipients.
I don't agree in this case where the system is designed to spread those posts. Fair enough if you don't want that to happen, in which case don't use a federated system where that is a design goal.
But for follower only posting, you are right, Activities go only to the recipients, theoretically to each individual inbox, practically to the instance's shared inbox.
While search is a problem, I think the bigger problem with the fediverse is expectations. People expect that their "follower" only posts are only shown to their followers. However you can get them anyway through so many other ways, its literally ludicrous that people expect anything they post to only be delivered to their followers. Boosting posts within Mastodon transmits them to federated servers, and this is by design.
My point has always been that if you play in a federated system, where you literally distribute your posts to anyone who requests them, you have to accept once it leaves your server you have lost all control. To then complain someone is getting that post is naive. That's the federated system working as intended.
If you don't want it public, don't distributed it via a public system.
I don't know if people were told otherwise, but they seem to be under the impression their content is fully under their control in the fediverse, when it is far from the case. Deletes/Edits of content are "should" in the spec, meaning yes people should do it, they are under no obligation to do so.
The only reaction the fediverse has to anyone they dislike is to block, and in some cases harass people till they leave. That wont work long term, and it isn't working now. Despite being blocked by hundreds of instances Mastinator is still getting thousands of posts an hour delivered to it, including posts from instances that blocked it because someone who is federated with those instances has not blocked mastinator and someone boosted it.
The current "elders" of the fediverse want to keep their behaviors and cultures. Which is fine. They should do so by having their own allow-listed federation, not accepting any follow request, or totally de-federating. They should not be bullying others into accepting their behavior. Not only does it not work now, its really not going to work once Tumblr, Flickr and others start integrating, as they will have enough people on that they will be the majority and they will make the rules.
“… It's a decentralized network that doesn't belong to any one party, so by definition there is no single culture on it. Different corners have different expectations and customs.” - https://mastodon.social/@Gargron/109323056922301691
Sure, and they should use Signal instead of publishing their conversations and then getting mad when they turn out to be publicly available.
The add the extra complexity of follower only things and instances and I just don't understand who this service is aim at other than the people who created it. Hence, to me, it is and will always be some niche thing (that isn't terribly interesting to me).
All that said I still read these threads because there needs to be a good twitter alternative.
Android malware is a thing and Signal Desktop does not encrypt anything at rest on disk.
This is one of those things that tech needs think is pointless when in reality opting out of search indexing drops your visibility to nothing which is good for making sure your posts don’t “break containment.”
Unless you’re a public figure or social media influencer having a post tied to your IRL identity blow up is usually a bad thing.
Well... Then be careful with what you post tied to your real identity. I have separate accounts for separate use cases. My main on Reddit, for example, can (very) easily be doxxed so I hardly post anything controversial on that account.
Through all the recent data leaks it's getting even easier to dox people based on screen names too. I'm even considering getting a PO box next time I move house, just to decrease the chances of my home address ending up in a data dump.
There's a preference in settings, but it just adds your profile to robots.txt. Meaning that if someone boosts your post and doesn't have that configured...
I certainly spend more time interacting with people in small group chats than I do on Facebook now.
The fediverse is a public protocol, and indexing public content is fair use(ML is not indexing, its very different), so any kind of licensing won't matter.
Discovery is already crap in mastodon and the fediverse in general, please don't stop people from improving it.
It may look like eternal September for fediverse natives, but that's what going mainstream is like, don't like it? Create something with some kind of intentional barrier like HN to stop the masses from joining.
I find the fear about getting doxed and having your words watched and used against you from fediverse natives very funny, because I guarantee you that most of them support cancel culture.
In general I find old fediverse communities too fragile, one example are the crazy content warnings(just mute words and topics you don't like, simple!) and now this.
The other bad pseudo-privacy idea is time-limited posts ("stories" or snapchat or whatever).
In both cases, you're crippling the software to add an illusory safeguard, which doesn't actually stop bad actors from having access to your posts and hence the ability to record and rebroadcast them.
You know who got the system right? Facebook. Private by default, but the user can decide exactly who can see each of their digital objects. It's easy to define groups of friends, allow sharing to friends-of-friends, one person, the whole world, whatever. The concept of friends (bilateral agreement to share information) makes way more sense than this "follow" thing.
I want my open-source, federated, Facebook already.
Edit: Although, I did just have a flashback to when Facebook announced Graph Search... which lasted about a week until searches that actually worked were deemed creepy and they backpedaled into the stone age. It's so frustrating how these technologies succeed or fail based on fashion rather than technical merit.
They exist... But it feels FB has gone out of its way to hide and obfuscate them.
Is my experience weird?
---
I quite liked G+'s focus on choosing your audience. Because you have sub-groups even within small, tightly-knit friend groups; when you raise that number into the hundreds it's only more true, not less. It was a mostly-effective UX for embracing that, and it led to my feed being dramatically more relevant.
Mastodon is filling a similar purpose for me, lately. The server you join has a pretty powerful impact on your local timeline - join a couple, use them as targeted sharing / browsing groups, and it's working much better for me than any algorithmic sorting ever did.
Each individual having their own personal view of their circles meant that you couldn't reliably know which of your friends had seen the stuff your reading.
That makes it really hard to talk about (Hey did you see X?, no what's X? Oh... oops?).
I like the theory of being able to organise my relationships into nice little buckets, but that's absolutely not how social things work.
"Did you see X" is largely killed by algorithmic feeds though IMO, which makes it somewhat irrelevant for any full-scale heavily-used network. Facebook is a prime example - important updates frequently are not seen by many close friends, because Facebook chose to not show them. Assuming nobody knows anything specific has kinda become the norm, sadly.
[1]: Obviously many will not, but that's fine. By following them you just get an unfiltered stream. But many of my friends couldn't care less about what programming language of the week I'm looking at (because they're not techy), or what nearby events I'm going to (because they're 1000 miles away) and I'm very much the sort of person who will categorize that for them so they aren't flooded with things they won't be able to join in on.
Hashtags are kinda like a crappy in-band version of this, and I have yet to see a system embrace them for this purpose. They're basically always for public purposes, which is part of why you need to use a million near-identical ones to actually get good coverage.
And then the best feature of all, Sharing of curated circles. A kind of competitive marketplace of topic related circles emerged so you could find these amazing circles of photographers or musicians etc. Best of all you ingested and then owned that (instance of) circle.
I will say though, it's specifically the old Facebook mentality that made me hyped back when Google+ was announced. And then it crashed and burned - which was exactly the point in time where I stopped wanting to be on literally any new social network.
> Within bonfire, you now have the possibility to define circles and boundaries: a way to privately group some of your contacts and then grant them permissions to interact with you and each piece of content you share at the most granular level.
>Boundaries go beyond the typical permissions on social media (i.e. who can see your content) and include a long list of verbs in order to represent all kinds of meaningful interactions and collaboration that should be possible on a real social network.
You could try and encrypt it down to the user level but a person added to a group would only see the content added from that point forward (the ability to decrypt would be defined at the time of the post, and new people could never read it).
The only way around that would be centralised key management which defeats the whole point.
This is one of those problems where it's probably better to just use Facebook.
Struggle here would be that anyone who gets the shared key would have access to view and potentially add to the group.
That one addresses a real threat-model - someone who was your friend and whom you trusted at the time you sent the post, but has since turned against you. That is quite common among teens, which is the main demographic these social networks are targeting. Rather than them having a permanent record of all your online interactions since you met including embarrassing photographs, it just becomes he-said-she-said like it has been for ages, unless they were actively subverting the program to archive content while you were still friends, which is rare.
It would suit some, but since this is federated, there are already instances with a search like qoto.org with a full-text search.
It's really odd that they made "no index" checkbox, but it's not cool to index. If they thought it is unpopular they should have made indexing opt-in not opt-out.
Going forward this will be instance specific thing, a lot of people want to be able to do searches.
At the end of the day, it's a federation protocol. The protocol's creator can have an opinion on whether things should be indexed for search, but it's up to users of the protocol, not the protocol's creator, how it's used.
(... and I have to comment on this bit from the article. "The problem would be a public search engine that Gamergaters and Kiwifarmers use to hunt down vulnerable targets." This heavily implies that Gamergaters and Kiwifarmers can't just build that for their own purposes, which would, of course, ignore robots.txt settings).
The choice of bogeymen makes me suspect the motivation is insincere, though.
Semi-public semi-searchable discourse is really quite nice when your aim is to not have the spotlight on you. Because there are unfortunately people that actively search for people to target and being on Discord makes that in practice impossible.
Bad actors have ignored them from the start, and are good enough at faking headers/using multiple ips such that they are very difficult to block
In the impolite space, these problems are addressed by encryption and auth/auth, not polite protocol. Which defeats much of Mastodon's designed goal IIUC.
Mastodon has a "Opt-out of search engine indexing: Affects your public profile and post pages" setting, which probably inserts a "robots" meta tag. [1]
But this is just for ordinary web crawling, and the bit isn't sent along with your posts that get federated to other instances.
[1] https://developers.google.com/search/docs/crawling-indexing/...
Another thing -- noindex is not federated. I used to enable noindex, and still found my content on Google (mirrored on other instances that I was federated with). I just allowed indexing.
I think Mastodon makes the correct call (everything here is public, because it's impossible for it not to be and still have the service be what it is), and the community wants a square circle. Yay activists.
The point of the service is to "spread the things you say most everywhere." That is the design. It's literally designed to do what the opposite of "privacy" is.
And yet, here we are.
I'm sorry, but these kind of discussions to me sound like "What if we could have email, except your posts don't go to any other person, you just read them yourself?"
I mean, you could use gmail to do this. It would technically work. But it's not what it's designed for, and much better ways to do this already exist.
Because that's the design.
If you don't what your posts to go everywhere, use something that's not designed for your posts to go everywhere on. There are other things.
I use search to do research on links all the time (https://www.swyx.io/twitter-metacommentary). If I read something good, usually plonking it into HN search or Twitter search yields a dozen more related points and rabbit holes I can go down. I can even engage with the author or find their thread of thinking or responses to a question I had that may already have been asked (or better, questions I didnt think to ask)
Without Mastodon search all these metaconversations about topics are lost.
Which is why over there I'm presently unpopular with lots of people on many sides of issues, because (this is rough, but hey) e.g. too many white people are going "ooh, no racism discussion here please, I'm too delicate." and also too many black people are like "if i see the slightest thing that looks a little bit like racism defederate and block that whole server!"
If Mastodon gets content licensing right, you'll still be able to ignore it and go ahead and crawl data when the license forbids, it scratches your itch and you're ethically challenged. But then if you do anything with that data in public you're going to get legal nastygrams. That may not even stop you, but it will drive up the cost of your lack of ethics.
Ask any security pro. You can't ever stop all the attackers. All you can do is make it more and more expensive to do bad stuff, and eventually most of them won't have a strong enough incentive to pay the price.
There are plenty of people on Mastodon - the vast majority is my bet - who, when there's a choice of content licenses, will cheerfully say "make it public", and then there will be excellent full-text search.
Are you saying there is currently confusion as to whether a Mastodon user inadvertently issues a license for their copyrighted content to be included in full text search, simply by using Mastodon?
If not, what is preventing someone from sending a legal nastygram now, given that no such licenses currently being granted?
Or are you saying that Mastadon users are not able to legal prevent indexing based on copyright alone (i.e. fair-use, or not substantial enough to qualify for copyright protection), and thus we need to force followers into some kind of private contract that they would break?
Similarly, just because you can't stop people from indexing Mastodon physically, doesn't mean that you can't stop them by social or legal means. However, I would add that the internet is really hard to control because of how open it is, which is why we patch security vulnerabilities instead of only relying on publicly shaming or arresting malicious actors on the internet.
Then everyone defederated my single user instance, as if that would stop an unauthenticated web crawler.
The idea that you could publish something on the web and expect it to stay private is insane.
...or you know don't say stuff in public if you don't want it to be seen by others. This is - and always has been - Internet 101 stuff: assume that the internet never forgets, and don't say anything publicly if you'd rather not see it on the front page of a newspaper.
I guess each new generation needs to learn that there are bad people out there, and computers make finding a needle in a haystack trivial.
You could've tweeted something 10 years ago with an anonymous account, and called it a day.
Yet, after more than a decade it can crash your life because you got caught by stylometry analysis.
This certainly applies to HN even more.
Correct. This is something I and several others have been doing for some time now. We have a private search engine that covers most of the Mastodon fediverse (including widely defederated instances), and there's nothing anyone can do about it.
Eventually, we'll give this data to the Internet Archive or put it in a torrent or something. It includes a decent amount of now-deleted content too.
Really, it's no different to what others, e.g. Pushshift, are doing with other sites. Except there's no opt-out. Anything you've already said is almost certainly in our data set, and it's there permanently.
> That’s extremely hard to prevent but isn’t really the problem: The problem would be a public search engine that Gamergaters and Kiwifarmers use to hunt down vulnerable targets.
Also correct. A couple of KFers we know already have access to search the data we've collected. It points them towards interesting posts or accounts to archive. People say some pretty wild things when they think they're not being recorded for posterity!
- permanently archiving posts they know people don’t want them to
- gave early access to known trolls
- is now publicly crowing about how there’s nothing you can do about it
- on a throwaway, because like most bullies they’re cowards
Personally, I’m left wondering if there needs to be an organized crime investigation into KF, for organized harassment campaigns which violate local stalking or other laws, and the role people like this account play in orchestrating it.
I think there’s be a certain irony to HN’s record of this comment being used to prosecute an accomplice in organized crime.
They’re on a throwaway because they’re too cowardly to admit who they are and face people like myself who want to hold them accountable for their bullying.
I think they’re exactly hiding who they are.
I act in a way I can publicly own.
You act in a way you know that you have to hide — because you’re aware that’s it’s anti-social.
If you go around a bar with a tape recorder and directional microphone capturing conversations from unsuspecting people and then use those private moments to launch protracted harassment campaigns, you’re an asshole.
No matter how much you say “they didn’t have an expectation of privacy!”
If you want to talk privately - talk to people that you trust and use apps like Signal that delete messages after some time.
What specifically do you feel is different?
It's completely different if someone sneaks into your home and leaves microphones there.
I think it’s very telling you made a strawman rather than respond to that point.
That’s anti-social behavior and it’s okay to call it such.
During the whole Cloudflare banning thing I remember coming across a dox that deanonymized someone via the Patreon hack dump [1], where a simple Twitter username match turned into a name + address (via credit card details stored by Patreon).
Anonymity and OPSEC requires some serious effort and knowledge.
[1] https://www.christianpost.com/news/patreon-hack-almost-14-gb...
The author implies GG/KF/4chan/whatever are incapable of writing their own scraping tools specifically for the purposes of harassment.
Case closed.
Making the information public kind of is the point of putting it on the web on a public url without authentication. Don't do that if you don't want it. And think twice before you publish something you are not comfortable with sharing in public. Not that hard.
But yes, it's a matter of time before these things get crawled, scraped, etc. by all sorts of media organizations, marketing companies, etc. Once you have a critical mass of people who are relatively well educated, with disposable income, etc. any self respecting spammer, advertiser, marketing person, etc. is going to want to be all over this. The fact they haven't shown up yet (at least not a lot) has nothing to do with access controls, just with a lack of critical mass. The fediverse just wasn't that interesting until people started showing up a few months ago. Now we suddenly have people like Tim Bray and other tech influencers showing up. The more clued in linked in junkies are already getting mastodon accounts. That kind of is the point of a good public forum. These people will want to be a part of it.
There are technical solutions of course but they are going to have to involve making mastodon more like a federated signal/telegram, which does not seem to exist just yet. Mastodon just isn't it. Encryption is what you need if you want things to stay private. Of course it doesn't prevent anyone with their own modified client making some modifications that archives the plain text after decryption and information leaking that way. But it would be a lot harder to scrape everything. You basically have to infiltrate every group you want to scrape.
Mastodon with some crypto added would be nice though. The mistake with email was that pgp usage never caught on (too complicated to manage for most people to bother). Signing message content would be a nice start for mastodon. It would allow people to build reputations and verify that messages are coming from who they claim to be. Impersonating people is a thing on Twitter. Without message signatures, it's going to be a thing on mastodon too. And the nice thing with reputations is that they are a great basis for filtering too. If enough servers flag a particular public signature, they might just decline to accept content by that particular signature. Or accept it but filter it out of the public feed by default.
-I have a real ID account with my picture and link back to my domain web page. I approve the followers to keep it to mainly people I know and trust. I post mainly to followers only or post publicly rather mundane information that I would not mind coming up in a deep background check.
-I have another anonymous account where I am most active and can express political and other views publicly without worry. I tell some of my friends of this account to follow after they have found my real ID one.
Lots of discussion about what privacy means no discussion about the elephant in the room. Networks that make acceptable use a choose your own adventure for instance by allowing people to self host and set policy on their own server end up with some people using it for things that the rest of planet earth doesn't find acceptable. For instance I'm certain that common web frameworks and servers are used to promote things both odious and illegal but because that machinery is invisible to most users the blame accrues entirely to the criminal.
With Mastodon the branding makes it possible for the blame to accrue to Mastodon as opposed merely the criminal because people are more apt to understand Mastodon as an open source twitter as opposed to a tool like Apache. Whereas this problem accrues to Mastodon the tool is in no position to dictate how users use the tool the relevant governments are and if prosecution becomes common hiding illegal porn from prying eyes will be done by the users themselves. If Mastodon suggests not federating with servers in countries that don't handle this issue eg Japan then search will tend to show off less negative content and indeed countries can be shamed into handling such issues better.
Mastodon itself officially supports Elasticsearch, but it isn't turned on by default and the UI limits what can be searched. I'm not sure if that's for social reasons or just to make installing a server easier, but there's no "never federate to servers with ES enabled" setting. Friendica has full-text search as well, and I imagine there are others.
Machine-readable copyright licenses as the author suggests could be very useful for situations where legitimate for-profit companies start federating their services. Perhaps I don't want Flickr[0] to show advertisements next to my photos that I share from a self-hosted Mastodon server, for example. I could apply a license that forbids that and have a good chance of enforcing it in court. They're not a good fit for dealing with bad actors though. If somebody sets up fedisearch.ru fed by legitimate-looking public servers, they're probably beyond the reach of the courts I have access to.
[0] The CEO of Flickr has publicly pondered adding ActivityPub support.
This feels like maybe not an unreasonable _default_, but I certainly don't think it should be mandatory. If a user wants to publish something on the internet, it's reasonable for them to be able to do that.
Personally, I am unconvinced by the long-timer's argument. I believe that if you are against Fediverse crawlers ethically, you must be against web crawlers as well, and indeed I spent months without Google, DDG, or any other crawler-based search -- StackOverflow tags, Wikipedia and bookmarks were my friends. It was an enlightening experience, but I now hold the reverse position.
I don't use Mastodon (or Twitter) but it seems to me like Mastodon was never really intended to be a pure Twitter clone, and what "normal people" want in this regard isn't really relevant. What's the point in trying to emulate a platform that feels icky and doomed, anyway?
I think they are missing fair use. IANAL but it seems unlikely that you need permission to do full text search if you only display snippets.
Honestly the whole copyright part seems misplaced. Ask the movie industry how well copyrighgt has worked to prevent people from sharing movies. If the concern is malicious people violating privacy, copyright is the wrong tool. Although what the author is describing sounds more like a contract than copyright.
https://www.varunsrinivasan.com/2022/01/11/sufficient-decent...
Which companies just say NO to getting more profits, every company wants to expand so I don't see why mastodon should promise private access to content when it's really not.
We take it for granted now that the public-ness of online content is very nearly binary: it's either public, and therefore hyper-public, globally accessible, indexed, searchable, publicly archived forever, or it is private. But this binary is artificial. Why can't we have things in-between?
Perhaps it's easier to understand if we make a real-world analogy. In the real world, there are public spaces where people interact and their interactions are, in general, not recorded, not remembered, not noticed. I can walk to my local public square and scream as much as I like about politics, and only people physically present there can hear me. If I say something strange enough, someone might film me and upload it to the internet, but that's an active choice. Almost everything that happens in these traditional public spaces does not have the hyper-publicity of web content. Likewise, I can have a casual conversation with a friend while walking in a public space, and people standing close to us might hear, they could even join in, but it's very unlikely that conversation will end up in the historical record. By contrast, while something like Twitter bills itself as a public square, it is nothing like it: no public square has cameras and microphones permanently fixed on it that globally broadcast everything happening there and pick up even the faintest voices. It really stretches the traditional conception of publicity. Previously, only celebrities and politicians had the misfortune of history's eyes being permanently fixed on them; what does “public figure” mean now?
Meanwhile, the web's conception of private is also an extreme. Conversations that nobody else can overhear and join are the only other option.
So, isn't publicly fetchable, but not indexable, content a reasonable compromise?
Now, certainly, the fact that there is no technical barrier to indexing means this is, you could say, privacy through obscurity. But that is true of almost all real-world privacy! It doesn't make it meaningless.
I also must reject the idea that because things can be indexed and archived, that they should be, especially when it goes against an explicitly expressed desire by affected users. Is mass surveillance only a problem when it is done in secret by a state?
The fact bad actors can't be stopped is a strange excuse as well. The world we live in is held together by the fact that most people, most of the time, act in good faith, even though they know a minority do not. If people gave up on caring because other people don't, there would be no society.
Finally: the legal aspect of the blog is such a strange attempt at proposing a solution. The fediverse is a system run by volunteers and its enemies (to be crude) are also generally volunteers, and the harms involved are generally at the micro, individual level (even if in aggregate they can be big), rather than having big corporate price tags attached. Licences are almost meaningless because almost nobody is taking anyone to court. The law cannot solve social problems like these.
The problem is, the effort required to index the whole of Mastodon is basically trivial, compared to wiretapping even a few locations, in the man-power, cost and ease of avoiding detection. With that in mind, either you never share posts publicly, or you have to assume it'll be recorded somewhere even if you don't want it.
> Lack of full-text search on general content is intentional, due to negative social dynamics
I guess you could frame this as a privacy argument, but I don't buy that as the primary reason people were upset with "Fedisearch." They were upset because it exposed their communities of lolicon enthusiasts. Do they deserve privacy or is that more of an excuse to keep their content off the radar of law enforcement and out of reach of polite society?
I did find this 2017 GitHub issue [1] to be slightly less inflammatory and full of fruitful discussion about legal implications of caching strategies for media content on Mastodon instances.
[0] https://www.secjuice.com/mastodon-child-porn-pedophiles/
[1] https://github.com/mastodon/mastodon/issues/1847
* "It" being the wider fediverse, where the 2nd and 3rd most popular instances allow "lolicon" content (which is illegal in many jurisdictions). Of course not all of Mastodon is filled with this, but all it takes is an obsessive minority to poison the grapevine with a large amount of content. Also, you can expect generative AI to lead to even more of this stuff.
Email is also full of disgusting stuff but nobody is going to abandon email because it's "full of child porn", although we all know perfectly well that there's lots of that stuff being emailed around.
Well-run Masto instances are actually pretty safe and pleasant spaces these days.
From my time speaking to Fediverse long-timers, no, this is not the case.