Scaling Mastodon is impossible
lucumr.pocoo.org
lucumr.pocoo.org
The point of mastodon is not that big companies will join small volunteer-run instances. They can run their own mastodon servers, just like they run their own mail servers and chat servers (or, more accurately, they outsource running their own mail and chat servers to Google and slack).
The right place for ActivityPub to land is that businesses and institutions will host their own (or pay a SaaS offering to do it for them, like they do with email and their CMS today), individuals will sign up to shared services (like you sign up with gmail for email, or blogger.com to blog).
The problem is going to be holding the barbarians at bay. All the same stuff we have to do for email will need to be deployed: Community blocklists for misbehaving servers, IP reputation, ML spam detection…
And funding it will require cash which the adtech industry will spiral in and offer to provide. ‘Federate with our servers and we’ll pay you 1c for every message you allow us to post to your users…’ Server admins will resist for a while but eventually, the walls will crack.
Enjoy it while it lasts, this burst of old school volunteer-run internet. It won’t survive this eternal September. It never does. As the architect of the Matrix said: Denial is the most predictable of all human responses. But, rest assured, this will be the sixth time we have destroyed it, and we have become exceedingly efficient at it
Who was, by the way, an homage to Vint Cerf, creator of TCP/IP.
There’s also an open and active GitHub issue on mastodon discussing how to separate server hosting from domain name so you can point DNS at an existing instance and use it from that domain: https://github.com/mastodon/mastodon/issues/2668
Many thanks to my friend on twitter for pointing me to both of these: https://twitter.com/__jesse_li/status/1592006641897320448
>Mastodon by Teams (tm), please log in with your active directory credentials to see the latest posts from HR!
APIs are not protected by copyright.
Everyone always just starts a group chat instead, as the reply is always so small and hidden, and the card system makes it hard to manage more than one or two threads at a time.
Yes, and still nobody uses it
The comms team religiously using it, and the ‘what’s happening’ digest being on by default meaning posts are surfaced in to people’s inboxes regularly until they unsubscribe means everyone knows about it.
Eve Online fans literally cheer Microsoft Excel features at annual Fanfest - https://arstechnica.com/gaming/2022/05/eve-onlines-ms-excel-... /// https://news.ycombinator.com/item?id=31292593
So this has only gotten weirder recently. Microsoft decided to rename Yammer when inside of Teams to Viva Engage. Nothing about this branding decision makes sense. It's still Yammer if you open a link to the web version. It's still Yammer.com if you want to go to the public web version. It's just called Viva Engage inside of Teams and now Outlook. (To use an ancient dumb aphorism, apparently it is now called Yammer in the streets and Viva Engage in the sheets.)
On top of different brand name (for the same app!) there's also a different UI design. It's almost the most confusing you could possibly make it.
Whoever is at Microsoft in charge of these marketing decisions has either lost their mind or this is some sort of long con that doesn't make sense yet. (There's hints that Microsoft is trying to rename Cortana to Viva, and they did rename what used to be Cortana Insights emails to Viva Insights, all of which are also dumb brand names. This seems an effort play up that these various inside-Teams and inside-Outlook sub-services are some sort of "intelligent business assistant for business users only" to jettison the final remaining ties to consumer Cortana [RIP] and the fun gaming-originated name for something bland and soulless and not-fun capital-E Enterprise, sigh.)
Anecdotally, the rebrand so far seems to be diminishing use of Yammer inside of Teams at my employer. They had Yammer as a default app in the default sidebar of Teams but when the branding switched Viva Engage is no longer in the default list. There's still something of a "well, post it on Yammer" culture here and people don't even seem to realize they can do it directly in Teams anymore because Teams no longer has a "Yammer app" and no one knows to look for "Viva Engage".
This is all just really wild and feels a bit like Microsoft has lost some of its mind here.
The language makes a huge difference as you need to deploy the whole Ruby toolchain, and front it with all the hard choices Rubyists had to deal with for ages to make it kinda work! You don't make a distributed competitor of Twitter following its own wrong design chocies!
Mastodon is old, and it's a shaky foundation, and it should be rebuilt from scratch instead of disappointing the millions of people who suddenly like Dorsey more than Musk... although Dorsey is working on Bluesky, not Mastodon, or ActivityPub!
Might I suggest you take a deep breath and tell yourself ‘it’s ok, I don’t need to pick fights with random people over trivial things’?
> The language makes a huge difference
If a decentralized social network is a good thing (jury is out imho), guess it’s a good thing that there’s ActivityPub implementations in multiple wildly different languages then?
ANyway, just try to imagine the same spammy bots taking down all Mastodon servers when the super eager to leave Twitter move to the next cool thing... It won't be pretty! The Fail Whale will feel like a joke compared to what's about to come.
> How was Twitter any better before him?!
Before Musk took over it wasn't saddled with the $13B in debt he took on to buy it. Before Musk took over advertisers weren't fleeing the platform in droves. Before Musk took over no one could impersonate Eli Lilly with a blue checkmark and cause its stock to crash. There's no indication now that Twitter will do anything to stop hate campaigns and abuse when they arise (hate speech is free speech, after all, and isn't illegal in the US). I could imagine some kind of future for Twitter before Musk, but that's immensely more difficult now.
I don't think Mastodon is the solution, mind you. But it's no surprise to me that a lot of folks are looking for an alternative.
But presumably because very few people cares about that benchmark and what is supported depends on what they choose to add. You're the first person I've ever seen mention it.
For a simple reason: Performance on dynamic requests is a choke point only for a miniscule proportion of sites, and the federation that is a selling point for Mastodon ensures that any single Mastodon instance is far less likely to ever approach a size where that becomes an issue. There are certainly sites where I wouldn't run Ruby, even though it's a language I love to use.
But I'll repeat myself: The choice of Ruby is far down the list of issues with Mastodon.
Returning to the benchmark, it's also contrived to reference as a general measure for a web app. Anyone building actual systems using Ruby puts Ruby behind a reverse proxy exactly because we're fully aware you focus on avoiding hitting the Ruby app server for anything static or cacheable, and so the setups on that benchmark (there are at least four frameworks and a couple of dozen entries) are entirely atypical for Ruby app server setups (they all hit the web server hosting the Ruby app directly).
That's fine - they measure what they measure, but they don't measure a realistic web app setup with any relevance to Mastodon.
E.g. the vast majority of traffic for a Mastodon instance is read-only requests to toots and media that are trivially statically cached, and so the vast majority of hits to a Mastodon server set up to actually scale should never hit the app server at all.
I have a deep distaste for Rails, and think Mastodon is more complex than it needs to, and there's plenty to criticise Mastodon for, but you're barking up the wrong tree.
For any software that embodies an open and formally-specified protocol (in this case, the protocol is ActivityPub), any first implementation of it is a reference implementation (as opposed to a defining implementation), because at any point someone else can create another implementation, and then both implementations will have to deal with adhering to the standard, rather than either being able to bully the other into implementing things a certain way. It's not about whether anyone else has implemented a second client for the standard; it's just about whether they can, and what would happen if they did.
Examples of things that are reference implementations of open protocols/formats, where the protocol/format "came from" the reference implementation but then became its own thing, with other implementations: Memcached, Redis, IPFS; PNG, OpenDocument.
Example of something that's not a "reference implementation of an open protocol", despite seeming like it: Amazon S3. There are many other "S3-compatible" servers; but there's no open, formalized standard for what "S3 compatibility" means; and so Amazon can change S3's API at any time, and everyone else will just have to deal with that. (They won't — it's been the same stable API forever — but they could. And what they can do is the point here.)
Key to remember is yet. The internet started as DARPAnet, after all.
All of this also coincides with a massive migration to Mastodon v4 adding to current maintenance burdens.
So the short answer is: masto.host probably will open its doors again if it doesn't seem to be hosting too much of the Fediverse, and it will probably be at least a few more days if not a few more weeks before registrations open back up if they do.
(There are some alternative hosts to consider if you don't want to wait on masto.host.)
Finding a mastodon instance is a pain today. And "run your own" is not a great option. (It's a pain to run your own publicly accessible anything, these days. I will never go back to running my own email server ever again)
I have only heard the best about masto.host but they have closed subscriptions for now (which only speaks for them IMHO). I have personal experience with Ossrox and Weingärtner IT, they are both excellent and I can recommend them without hesitation. I have no affiliation with any of those companies except being a customer of the later two.
EDIT: The price of a small instance is less than a Twitter subscription. So, if you want your own instance - which as pros and cons - nothing is holding you back.
While the verification green check mark is nice, it is much better to have PR hosted on toot.company.com instead.
Furthermore, there's the "who can see what" - https://restoreprivacy.com/mastodon-privacy-issues-risks/
> Because these messages aren’t end-to-end encrypted, the server owner can read everything in plaintext.
(See also https://github.com/mastodon/mastodon/issues/18079 )
That's a BIG "companies should be hosting their own servers."
But does this really address that issue? I see this as an attack vector really. Let's say Elon isn't happy with the competition. He can join a small server and force them to either increase hosting fees or to shut down. That's really the point of the author's comment and the relationship to scaling.
Please correct my misunderstanding.
This is perhaps my biggest regret. I've ignored Mastodon because it seem too much like a PITA to deal with. Now I've been looking at it more seriously, but so has everybody else who's on Twitter, so by the time I get there it's going to be... Twitter pre Musk. This isn't really that appealing. Coulda woulda shoulda, etc., but I just wasn't paying enough attention at the right time. Rather like Bitcoin. Don't get me started.
Mastodon doesn't need to be Twitter where you "must" follow celebrities of any sort or streamers/podcasters/YouTubers/Influencers. People claiming that Mastodon is terrible because it isn't focused solely on massive instances with maximal "reach" are missing the point: that is a *feature* and not a bug. People have forgotten the early days of Twitter when it was mostly about following real life friends, or conference acquaintances. There's a usefulness to quieter, stiller social media. It's great to have tiny instances with tiny follow streams.
Also, on Mastodon you will only see what you explicitly follow. You won't even see the rabble unless you go looking for it.
It's not like torrents and likes of PirateBay that make downloading movies easy. It's very complex to get into the network. It's like Mastodon is trying to not let you in.
Not really. Email and ActivityPub are radically different in one way that makes all the difference: AP is whitelist-based (follows).
> Federate with our servers and we’ll pay you 1c for every message you allow us to post to your users...
And this is a fine model - much better than all the power concentrated in one stagnant entity that forbids everyone else from participating, extending or improving.
So now your user's incoming follow request inbox is a list of viagra ads, .eth scams, and pornstars.
I'm saying that in terms of economic "bang for the buck", if you were going to spam the fediverse doing it with follow requests is a terrible way to do it with very low likelihood of success. You are better off trying for other means of virally distributing toots than follow request spam.
Also, federation itself may be allowlist-based. Some instances already run that way. Likely many more would if inter-instance spam follow requests rose above a certain threshold.
I'm not thats anywhere near the disruptiveness of SMTP. As a private person, I could simply have the server not notify me of any follows and it wouldn't affect my own feed at all.
The two smartest people in this business are Yishan Wong and Ellen Pao.
Just spitballing ... offer a generous severance for voluntary layoffs now along with a hiring freeze, and give a mandate to simplify the offering and reduce infrastructure costs. Take the time to learn the business and identify which areas can be profitably wound down. Only then reorganize the company and make targeted layoffs. (Bonus idea: don't rush features like Twitter Blue that no one asked for, are ripe for abuse, and might even lose money overall if reports are correct.)
I've never been a huge Musk fan, but any respect I had for him as an effective business leader has evaporated.
What's the problem there, if you know all your users have corporate-mandated spam-blockers installed that will drop the spam they're paying you to forward?
Wil joined what was at the time my own home instance, following something of a flameout on Twitter, itself caused by 1) people harassing him there, 2) Wil employing a shared, user-generated blocklist which 3) by both its own size and celeb-driven adoption rendered many other Twitter accounts effectively blocked --- not by Twitter, but as a result of individual adoptions of that blocklist. How legitimate inclusions were, and/or how feasible it was to get off the blocklist, I really can't say. That said, Wil arrived on Mastodon with both a large fan and foe base.
Harassment of Wil began on Mastodon.
<https://www.theglobaldispatch.com/mastadon-shuts-down-wil-wh...>
The admin/moderator was ... less than transparent or capable, and responded to a case of "this profile is causing me problems" by banning that profile rather than seeking out the root cause of the griefers who'd in fact triggered the problem.
<https://mastodon.cloud/@wilw/100635779449174251>
Ultimate, Wil left Mastodon as well, writing his own account of the matter:
<https://wilwheaton.net/2018/08/the-world-is-a-terrible-place...>
Mastodon's evolved its own tools and practices (both technically and as a community, including a community of admins), and should perform better in future, though much engineering progress emerges through pain, suffering, or worse.
Wil is, in the world of celebrity, fairly modest stakes. I've heard tales of Twitter that there was a point where it could only handle a tweet from one major celeb (e.g., Justin Bieber, Katy Perry, or one of the 'dashians) at a time, and effectively had servers dedicated to sufficiently large accounts. Celebrities proved an interesting stress-testing mechanism.
Mastodon ... has not yet experienced this, and I've my concerns with how well it might handle such scaling, or what specific elements of scaling (users, instances, posts, interactions, ...) will prove most problematic. I've certainly seen and heard of numerous cases of growing pains over the past few weeks, though nothing insurmountable.
But:
1. Celebs and brands are profile types that should be considered.
2. Mastodon's had some experience and performed poorly, both technically and culturally.
3. Adding very-high-profile profiles to the network needs some consideration. It's quite likely that there will be poorly-thought-through adoptions in future, however.
... but then a bunch of guys in stormtrooper outfits start showing up every day and the trekkies start fighting with them and the cops are getting called out and pretty much eventually the owner is going to ask Mr. Wheaton to stop coming by.
And because it's just someone else's café, he pretty much has to do that.
The good news is, going to that café is not the only way for him to participate in the fediverse. He is the proud owner and operator of wilwheaton.net, and he can point that at his own activitypub server, that nobody can kick him off. He can bulk block all stormtroopers, interact with the trekkies in the café, and handle his own bandwidth and compute bills for delivering his content to all his fans.
Bonus: the fact he's posting from @wil@wilwheaton.net acts as a pretty solid cast iron guarantee of authenticity, that you are dealing with the one and only actual Wil Wheaton.
Meanwhile, Stephen King is sat at the café down the road trying to figure out why the new owner there wants him to pay him $8 a month to have the owner keep telling people who walk in 'yep, that's the real Stephen King over there'.
I've been on Mastodon since 2016. I enjoy and appreciate it. I admire the goals.
But much of the "it's intrinsically better / safe / scalable" talk seems based on little more than hope and faith, rather than evidence and engineering.
And given the Fediverse's own protocols, operating out of MyCafe rather than SafeCafe or BigCafe ... only provides modest additional protections. Yes, it helps assert identity more effectively (as I've previously noted: <https://news.ycombinator.com/item?id=33512169>), but it only solves a small portion of the abuse, trust, and safety concerns. Mechanisms simply are not yet sufficiently robust for a large-scale (100m -- billions of MAU) social network.
Some of that will come in time, though I've concerns that there are weaknesses that have already been baked into the underlying architecture and assumptions.
To my understanding, there were further complicating factors involved beyond just "scaling":
- At the time I heard that the mastodon.cloud instance admins and moderators spoke English as a second language (it was started to primarily focus on Japanese Mastodon, to my understanding), which added an accidental language barrier to the challenge of moderating that many reports (and being able to seek out a "root cause of the griefers")
- I still don't know how to make sense of the whole counter.social thing and Wil's supposed involvement there (which happened before the mastodon.cloud incident). I think it was partly a scam that accidentally honey trapped Wil with very bad timing, but he doesn't refer to it directly in his own account, only obliquely, so I don't think we know an answer for sure. (Only what I observed in federated Meta drama from a tiny instance at the time.) But to my understanding it still relates to some of the root causes of some of the grievances. (counter.social is a now entirely defederated instance with alt-Right associations and markets itself as "safe" and "secure" and "celebrity friendly" and "InfoSec focused" and is run by an admin referred to most commonly as "The Jester", seemingly somewhat in homage to/love for Batman's nemesis The Joker.) I can't imagine knowing the full facts ahead of time that Wil would have ever chosen to associate with that sort of instance on purpose, but I can understand (some) of why once accidentally associated with that sort of instance it caused some lasting grievances.
My understanding is that Mastodon.cloud's admin was a native German speaker. The instance was subsequently purchased by a Japanese firm. (I left during this process due both to dissatisfaction with the WW incident and lack of clarity on the transfer.)
WW had nothing to do with CounterSocial AFAIU. As the link I'd provided above shows, he'd been on mastodon.cloud at the time: <https://mastodon.cloud/@wilw/100635779449174251>
That said, I'm frankly not familiar w/ CounterSocial, though it appears to be a Mastodon fork:
- counter.social's admin "The Jester" was the "InfoSec expert to the stars" that was behind at least one of the post-GamerGate blocklists for the safety of celebrities that Wil had been recommending to other celebrities and because of that Wil had some sort of (para)social relationship with The Jester, at least on Twitter.
- In 2017 (nearly a full year before the mastodon.cloud incident), The Jester allegedly convinced Wil to join counter.social (or deeply faked an account for him on counter.social). [1]
- That @wilw@counter.social account went viral on Mastodon among instance admins because up until that point counter.social was only a weakly federated instance few were familiar with/"no one had heard of".
- counter.social going viral brought a lot of scrutiny on counter.social. It was quickly found to be full of accounts of transphobes and racists and worse.
- This lead to a public re-examining of The Jester's Twitter blocklists. Those blocklists were found to be themselves biased towards transphobia and subsequently celebrities were asked to disassociate with The Jester and stop using his blocklists.
- Most Mastodon instance admins I trusted blocked/defederated from counter.social in the middle of these events in 2017. (For what it was worth, I made the same decision and also blocked it.)
- I still don't think this association with counter.social was intentional by Wil. This part is mostly my speculation: It sounded like "InfoSec expert to the stars" The Jester had been trying to get any celebrity to take the bait to join Mastodon via his instance and it was a bit of an intentional honey pot trap from someone who had built up a "trustworthy" reputation on Twitter taking advantage of the consequences of GamerGate to build that reputation and only with hindsight seems to have been doing that "for the lols" and was pro-GamerGate while pretending to be anti-. (That era GamerGate of Twitter was a different time and wilder place. Hard to imagine people not having immediate paranoid reactions to someone calling themselves "The Jester" today, but hindsight is 20/20 and easy to armchair quarterback when not in the middle of the stress of it all.)
[1] The best receipt I have found so far is mastodon.social's ancient copy of the account profile and 2017 pinned intro post: https://mastodon.social/@wilw@counter.social (this is also what I saw on my own instance at the time and matches what my instance still shows; I don't trust counter.social enough to see if permalinks track all the way back into that instance's own slimy history.)
(ETA: Also, sadly counter.social has been in the Mastodon news again lately because new people in this massive wave of Twitter users looking for a Twitter alternative have been duped into thinking counter.social is a "safe" alternative to Twitter and not a Gab-like haven for trolls.)
Isn’t that a protection racket?
Hard disagree with this. Email is the example I give people of how federation could work. You can use any email provider and interact with any other email provider effortlessly. This is undeniably an improvement over internal direct messages within a centralized service (i.e. Facebook or Twitter DMs). If you leave facebook, you leave your contact list and message history. On email, if using a custom domain, you can switch email providers or even host your own server without anybody on your contact list even having to know this occurred.
Lots of people have been banned from platforms, rightly or wrongly. You can't get banned from email. You can't get banned from having your own website. You can't get banned from the telephone system. If you don't own your own data, you are at the mercy of those who do.
> On the topic of moderation the very same issue is even more absurd. Some instances want uncontrolled free speech where some instances effectively are pure shit-posting instances which are completely de-federated from the most of the fediverse as a result. Other instances really like to control their content, where some popular ones such as fosstodon ban all languages than English as a result to allow moderation.
These are talked about as problems, rather than being the primary selling point.
If you're on Google or spamhaus's blacklist, you're effectively banned.
Yes, a specific server or address can be "banned". But there is nothing stopping you from just opening another gmail / yahoo / whatevermail account.
From me as well.
> You can use any email provider and interact with any other email provider effortlessly. This is undeniably an improvement over internal direct messages within a centralized service
But the general public went for siloed systems instead.
So I’d say you or me want decentralization, but I don’t think that holds true in general.
And yet some former Twitter users have identified why this is a problem, and are now moving back towards a decentralized platform. It's why scaling Mastadon is even a topic right now.
That depends on how one looks at it. If spam and illegal content are considered, it may not be "working" in the sense a public service needs to be.
Email works because it's private, so the requirements are much looser; a large part of the article is how to solve the bad actors problem, which doesn't apply to email (well, it applies, but spam is more or less accepted as fact of life, while fake news etc. isn't - and it may never be, independenly of being right or wrong).
By that standard, centralized social networks have utterly failed us as well.
They're not on the same level; moderation of social networks like Twitter/Facebook can be considered insufficient, but it exists. Contrast to the very low barriers (if any) to send spam emails, where the filter is, ironically, in the hands of centralized services like Gmail.
Except e-mail has been a total failure when it comes to spam/abuse, and in practice it isn't this federated, decentralized paradise at all -- virtually everyone I know uses Gmail for their personal account. And if you want to run your own server, good luck getting Gmail to accept your e-mails at all.
And if you can't get Gmail to trust your server, that's pretty much a ban for all practical purposes.
The point is that they can and do because the system is not, in fact, centralized.
Also, judged only on the volume of spam and abuse, many systems besides email -- decentralized and centralized alike -- could be called a "total failure." And yet somehow these systems remain stable and functional and useful.
After using the same Digital Ocean IP for 8 years, I got unlucky and their /24 block got listed on spamhaus (or mailcop, or similar) and major providers (Outlook and Gmail) stopped accepting messages from my server.
It works great until it doesn't
The benefits outweigh the costs, but it is not easy, and you can start having problems that are very hard to solve.
I've been using my own domain for my email for over two decades. In that time it's been self-hosted, hosted with Google, and hosted with Fastmail. I control it, so I can take it wherever I want. I've been able to take my full message history with me too.
It has not been a "total failure" with spam/abuse either, though I'm glad to have Fastmail handle that for me these days.
In any case, I'll take the tradeoffs over a centralized system.
> though I'm glad to have Fastmail handle that for me these days.
I get less spam in my Fastmail account than on my remaining Gmail accounts, but far more importantly for me: I get far fewer false positives with Fastmail. The amount of real mail that used to end up in my Gmail spam folder made the spam filtering pretty much pointless because I ended up going through the spam folder daily anyway.
Same here since the end of the 90s. Completely agree with the rest of your post.
I don't believe that's true:
https://blog.maartenballiauw.be/post/2022/11/05/mastodon-own...
Also https://masto.host/how/ and https://simonwillison.net/2022/Nov/8/mastodon-is-just-blogs/
i think we should have learned by now that when “some people are saying…” we should be very skeptical if those claims are based in reality.
in this case, they’re not. if you wish to move from .social to .net you move your account and your million followers go with you.
If I understand correctly, /.well-known/webfinger should do the trick.
I’m “at” the same long term domain in several protocols—
@name@mydomain # ActivityPub
@name:mydomain # Matrix
name@mydomain # Email
Beyond that, Twitter is rife with spam, scams, and tons of other abuse. So is Facebook. So is reddit. The fact that email doesn't work that way doesn't matter.
So the net number of spam emails in my email inbox is less than 1 per week. On Twitter, for me it's at least 1 per day and there's nothing I can do to tweak the filtering.
It's a total nightmare out there!
I have clients on big Hosts like SiteGround, and sending email is a problem!
I'm 100% in support of the idea that services should be protocols though! :-)
Which, ironically enough, is the answer to everyone’s problems with mastodon.
Does everyone you know use gmail for their business accounts?
Does absolutely, 100% of the people you know use gmail for personal accounts?
I’m going to go out on a limb and say no to both the above.
There’s no reason mastodon has to be everyone runs their own server just like everyone doesn’t have to run their email server today. People use the company server for work stuff and fluffycat233575@yahoo.com for personal stuff.
I don’t know if it’s just the HN bubble but all people post is why it can’t (or shouldn’t, because, get off my lawn) work. It’s like people have a vested interest in making sure non-adtech doesn’t succeed, almost like their very livelihoods depended on it.
—edit—
And, back in the day, spam filtering was something you had to do on your own or just dealt with all the spam. The big email providers became big in part because they solved that problem for most everyone. In ‘97 nobody (OK, nobody but the US postal service) was saying there needed to be this gigantic corporation monitoring everyone’s emails for wrongthink or only a central entity could solve the spam problem. People just bucked down and solved the problem enough to get a sizable portion of the population using email…I’m genuinely curious if there’s ever been a technology with higher uptake than email now.
Yes, and the major way in which they did it also made it hard to self-host an e-mail server.
> I don’t know if it’s just the HN bubble but all people post is why it can’t (or shouldn’t, because, get off my lawn) work. It’s like people have a vested interest in making sure non-adtech doesn’t succeed, almost like their very livelihoods depended on it.
Perhaps it's true in some cases, but I think it's more likely that people here recognize the new federated communication services are structurally equivalent to e-mail, so there's little reason to expect they'll end up in a better state than e-mail did. On the contrary, there are reasons to expect they'll end up worse, because e-mail had the benefit of predating the commercialized web and mass adoption of the Internet by not tech-savvy people.
The issue here isn't spam, though, it's about who sets the rules. Spam is just one of the strongest drivers that make a federated network evolve into an oligopoly. It was the driver in case of e-mail, but e.g. for Mastodon, I'd say political extremism is the primary force. The result ends up being the same, because the dynamics are content-agnostic: users migrate to nodes that offer most reliable communication, and as those nodes grow in usercount, it makes it easier for them to fight undesired communication with blunt tools (e.g. instance or IP blocklists). It's a positive feedback loop.
And they show up on mastodon, because everyone is talking about it, and find it isn’t the safe space they carefully curated with their twitter blocklists and write a blogpost (because they can’t complain on twitter anymore) about how it’ll never work.
I don’t understand why these tech-savvy folks can’t give positive criticism to maybe fix the problem instead of “it’ll never work, can’t we just have our twitter back, pretty please?”
All the filters and whatnot people talk about people talking about Musk dismantling.
There's a mass hysteria happening. Pretty much nothing of consequence has happened on Twitter on user side, except for what people make it in their heads, or talk about to show off in front of the group, one-upping each other in their newly discovered deep hatred for Musk.
They check out Mastodon and are disappointed, because despite the fact it does give you more control and tools to "carefully curate your safe space", it's not the "Twitter but without Musk" they hoped it to be. It's not, and can't be, because the userbase isn't there, and the culture isn't there. Fediverse is still driven by political extremists crafting their own safe space to stew in their own vitriol, and the very identity is still defined as "we're not Twitter". Fediverse won't grow until it develops its own identity, that's not just negative space between commercial social media (and until it insulates the federation against political crazies).
Federation is hard. E-mail and Mastodon are both, simultaneously, large successes and huge failures. There's a lot to learn from it, there's a potential to do better - but now is not the time, not when the topic is covered by the wannabe Internet stars doing performative anti-Musk dances.
> “(...) can’t we just have our twitter back, pretty please?”
They still have it, and guess what, they're still on it - because all this whining is just for show.
You mean the pay to verify thing never happened?
The new boss has a different opinion of free speech than the old boss, (and who he's allowed to cancel) and is starting to make changes effecting that point of view. Downplay that all you want, but that's something of consequence for a social media site. If Twitter were a car company, it would be difficult but not impossible to ignore all the talk and focus on the product, the cars. However, seeing as how Twitter is "only" good for talking on, the tone of that conversation changing is quite important to the people in that conversation, and they are liable to have gasp conversations about that change.
[0] https://twitter.com/amymek - banned in Germany for Nazi shit
[1] https://twitter.com/realadolphhitle
[2] https://twitter.com/funnymonkey9876 - "dead tw*tter employee hanging from a tree groyper" Their pinned tweet is especially lovely. It's been up for almost 6 hours. (cw: boringly uncreative use of the n-word) https://twitter.com/FunnyMonkey9876/status/15735249917940858...
[3] https://twitter.com/richardbspencer - Nazi that previously lost verification for throwing Nazi rallies.
Your politics and their relation to the above, and whatever you believe of free speech; it's uncontrovertible that it's no longer their Twitter, it's Elon's Twitter now, and not everybody is going to like that.
The second hand also has to follow spammers. Most humans just don't do that, do they?
> It was the driver in case of e-mail, but e.g. for Mastodon, I'd say political extremism is the primary force. The result ends up being the same, because the dynamics are content-agnostic: users migrate to nodes that offer most reliable communication, and as those nodes grow in usercount, it makes it easier for them to fight undesired communication with blunt tools (e.g. instance or IP blocklists).
I agree with this bit. But I think there is an opportunity to avoid it this time. We're much more aware of the potential pitfalls than when email was conceptualized as a protocol, and we have some quite amazing tools to tackle some of these challenges differently. I predict we'll see some innovation on this front, and I'm honestly pondering participating in that effort. The fruits of that would be very much worth it.
Regardless, I don't think it's contradictory that people flock towards a somewhat-centralized system like Gmail; the point is that it still supports an open protocol. The fact that Protonmail "just works" when emailing Gmail is a testament to the fact that federation can work in practice, even if it does start becoming more centralized.
I haven't been able to send mail to Pocket support for about a year based on Gmail blocking.
YMMV.
https://cfenollosa.com/blog/after-self-hosting-my-email-for-...
I still can, but only because I hate myself enough to run my own email for more than 20 years, so there must be some magic field of trust somewhere in the background. I wouldn't start from scratch though.
Less flippant, very few companies except startups and one person shops use GMail in Germany. I'm no fan of Outlook and Exchange, but that doesn't mean I can just ignore it away.
In other words, Gmail is the most popular email provider because it's in Google's interest to have people be logged in all the time so they can personalize search results, so that incentivizes them to make a very good email system and then give it away for free. Giving something good away for free because you gain indirect benefits via some other business will rapidly centralize more or less anything, which is why there are at least theoretically rules against market dumping and tying (which aren't really enforced in the software world, but that's another matter).
We can imagine a parallel universe in which search is far more competitive, with lower margins, and thus Google couldn't financially justify subsidizing consumer Gmail for so many years. In such a world it's likely that there'd be more players, perhaps they wouldn't be as good but there'd likely be more of them, there'd probably be companies that specialized in selling spam filtering tech to them and it'd look more federated than what we have today.
Freedom doesn't mean the requirement to spend the community's resources propagating material that the community agrees is harmful or objectionable. That would be the opposite of freedom.
The ability to refuse actors that the community doesn't want to spend resources on is essential to freedom.
Every actor has the freedom to choose whether to conform to community norms, or to form their own communities with different norms.
Your idea here works when the system is fragmented enough that these choices are personal choices, but the existence of so many large platforms makes their choices systemic, not personal.
On top of that, there isn't community agreement as much as there is community ignorance. For every case we hear about of someone being banned or blocked by an automated system despite doing nothing wrong, there are hundreds more similar cases which simply don't gain enough traction for the 'community' to hold the service provider responsible.
It's fallacious reasoning to assert that any system that is less than 100% perfect should be changed.
If the current system meets the community's expectations 98% of the time, and alternative systems meet the community's expectations 90% of the time or less, it would be silly to throw out the current system.
That's like saying stop signs at intersections should be removed because sometimes I have to stop when there are no other cars there, and sometimes other drivers fail to stop when they should, and some stop signs have been defaced or knocked down.
It's amusing that the first response to that point being made is to miss it, and to somehow paint the isolation of smaller servers as a strength of federation (because it gives users control over what they see.) Then the next one is literally an argument that decentralization will inevitably centralize, because that centralization will satisfy the communities expectations 98% of the time.
The last paragraph wanders very close to claiming that it is childish and entitled to expect your decentralization not to be centralized. Doesn't that just enthusiastically reinforce idlewords's claim?
Even putting aside the slightly absurd difference in stakes with my example, there's also the issue that you're just arguing for centralization, which defeats the point of a system designed for federation in the first place.
The new orthodoxy requires us to assume that all censorship is in good faith. "What were you doing to make them censor you?" is the question of the day.
And the answer increasingly often is, "I did not censor the other people quickly enough and visibly enough when I was told to."
Like when you go into a pub and start yelling insults at the guests, the patron will "censor" you by throwing your out of their private locality.
The concrete truth is: majority of users will do whatever to make their e-mail (or other form of communication) work, so the easy solution to spam is to blackhole anyone that isn't coming from or vouched by a major provider, and have the minority with specialized needs jump through hoops to get unblocked (or preferably, to change their needs/preferences so that mainstream provider is sufficient).
There isn't any actual user community here. GMail address owners aren't a community. Twitter users aren't a community. Mastodon users aren't a community. The only community is that of the federation instance owners, and it operates as oligarchy. The major players decide who they'll peer with, the minor players have to accept it if they want to be peered with. The actual, non-crazy-activist users, have no power, and are forced to either use the "blessed" providers and accept their rules, or not use the service.
This is already the case for Mastodon for quite a while, BTW.
* People don't need nu-Twitter to be perfectly reliable they way they do for mail
* For that matter, nu-Twitter doesn't really need to be a unified namespace; it can be less coherent even than email addressing and still add value, the way RSS feeds did.
* In the years before Google centralized all of email, it was pretty common for people to have multiple email identities on different services, which seems like the worst this is going to get.
I don't love the software, and the user experience definitely isn't ready for prime time yet, but the basic model seems workable?
He doesn't even mention email or phone service, both of which are decentralized and work well enough.
Except it isn't actually decentralized in the same sense as Mastodon is. You can't self-host a bank.
> He doesn't even mention email or phone service, both of which are decentralized and work well enough.
The same with phone service - it's not decentralized in this sense. You can't just run your own phone network, you can only choose from a single-digit number of major providers in your country, which are increasingly just multinationals.
The only reason banking and phone service may sometimes feel they're decentralized is because of the high interest governments have in them, and the copious amount of regulation in the industries that prevent the worst kinds of consolidation and abuse of power.
Email is, indeed, decentralized. And as many commenters already wrote, it's a disaster and a great example of how federated systems fail.
That presupposes that federation is actually a goal and not a means to an end. It's amazing how great developers are at not understanding what creates value. Twitter's value is _reach_. Get 400M users in one place, hook them with interesting content, server them ads, nudge them to sign up, track their engagement to make the platform sticky and attract more contributors. Any censorship that has been done has been for one and only one reason and it's protect the value of the platform. Primarily to advertisers, but also to users because users make the platform valuable. Decentralization deliberately cripples the main value of the platform (reach) in exchange for what exactly? Lack of moderation?
If you want decentralized, uncensorable communities, we solved that 30 years ago with usenet. Or, like you said, you can't be banned from email or running your own website. That's always been the case. People still write blogs. Yet Twitter sold for $44B and Tumblr was sold for $3M. Why? Because Twitter has reach.
The platform may not be able to automatically enforce a ban, but you can still be banned via non-technical means. Not using email or the phone can be a condition of bail or parole, for example, like how Kevin Mitnick wasn't allowed (by court order) to use the telephone system in the early aughts.
If there's anything to be learned from the KiwiFarms saga, it's that these statements are no longer true.
This saga has really muddled up responsibilities of different parts of the chain that makes the internet, where companies that should not be filtering traffic are doing exactly that.
This right here is why I can't take anyone seriously who takes the free speech "issue" surrounding corporate platforms. There is no free speech on a platform, ever.
The problem starts here. Not many people who are not tech savvy even know what a custom domain is. Let alone having an email with a custom domain.
> You can't get banned from email
Ask people who get locked out of their gmail account.
What we techies miss is that there are more people who don't understand technology and are not willing to spend time learning about it as they have other more important things to do. Tech is just one of the tools they are using to get their tasks done. We cannot expect people to spend time understanding everything about tech. It is not surprising that it took a commercial company with aggressive behaviour when it comes to controlling users data to put BSD on regular non-tech consumer desktops. And why Microsoft Windows succeeded in being on regular consumers desktop OS whereas commercial Linux-based OS struggled; and the company that successful put Linux on mobile devices also has quite a record when it comes to users data.
We techies can talk a lot about decentralisation and owning data. But the regular non-tech users really don't care. They just want simple things done quickly using tech and get on with their lives.
Recently I saw videos from multiple news channels on how to use Mastodon. They were at pains to explain to users how there are multiple servers and one has to choose a server, which is run by individuals or organisations. And if the server goes down then the user has to move to another server. Regular non-tech users really don't want to get into this complexity. Has anyone ever seen TV channels explain with great effort how to use Twitter?
I would point out that the concept is really no different from how ISPs work, and yet the public seems to have wrangled with that one.
Geographic locality helps. At any given location, the choice of ISPs will be severely limited. If my experience in a major European city scales, there will be between one to three ISPs able to service your areas, plus a few more niche/specialized ones. The actual choice is usually made based on ads, or asking a tech-savvy acquaintance for help.
I get that choice can be overwhelming, but that generally doesn't change whether people get the concept.
If someone could figure out a way to provide internet to consumers so that they didn’t have to deal with ISPs/carriers they’d capture a huge market overnight.
> Ask people who get locked out of their gmail account.
And, from the other end, ask people who actually self-host e-mail - not just own the domain, but also run their own mail server. The anti-spam measures of major e-mail providers are, in practice, banning insufficiently determined self-hosters from e-mail.
I've been doing for 2 years now. Mailinabox. occasionally using a one line terminal command to update the system but other than that, smooth sailing.
my only gripe. domain buying/administration should NOT BE THAT HARD for beginners.... its 2023 almost and setting up a domain is very difficult for new users...
i excensively use email alises and filters, something not possible in gmail or others (yahoo mail has that for free) and just that has made the whole thing worthwhile for me.
email aliases are charged an arm and a leg compared to other features but in your own systems, its great
iCloud makes it very easy to register a domain through Cloudflare and route email at that domain to your iCloud email: https://support.apple.com/guide/icloud/purchase-a-domain-mma...
It is impossible to use a VPS from any sizable or affordable VPS servers. The IP blocks from digital-ocean, linode, AWS, OVH, etc all are, have been, or will be randomly put on blacklists. Not the blacklists that you can apply for to be deregistered from. It's sheer impossible to have mail from a mailserver on any such host delivered to hotmail, gmail or such.
It is impossible to run a server at home. "domestic" IP addresses are treated similar to those of affordable VPS providers.
It is expensive to get a clean and proper IP address. It takes years of serious volume to give an address any authority. It takes only one mistake by any of the thousand typists at hotmail or gmail to undo this investment. Without any way to recover from it. None of these large mailproviders have any incentive to improve this: an "engineer" at their end making a "mistake" actually benefits them.
I've ran a mailinabox for as long as the project exists. But I'm sunsetting it moved everything to tutanota (but don't like the service so will move to an alternative in january). I just gave up: the monopolies have won.
the situation isnt that bad that you make it to be....
i;ve been using them for like 2-3 years now and have had no problems. these are yearly prices so yeah
I also know how to bake my own bread, but I'd rather get it from the store.
Completely self-run email is a bad user experience, just like making my own bread. Most people here forget that people want convenience over anything else.
I'll go stronger than that: if your theory is that people should want to be their own sysadmin, and people who don't want to be a sysadmin don't deserve [a thing] then your theory is bad and you should feel bad.
This problem is so bad that's it's driven a quiet de facto re-centralization of email.
Usenet is another example of the problems with 0 moderation and federation. Sure, it distributed amazingly (many ISPs mirrored usenet data), but once spammers started spamming, it quickly died as a discourse medium.
Those with concerns over censorship need to face the fact that we've previously had 100% censorship free platforms and nobody uses them anymore. We dumped those platforms because of the lack of censoring, not in spite of it.
Heck, HN is moderated as well, I don't think most of us would be here if it was unmoderated.
How does spam work on a social-media-like system where you have to wilfully follow someone to start getting their feed? Do you volontarily subscribe to receive spam?
> All it would take is popping a poorly configured server and then spam can be injected into the feeds you've chosen to follow.
This is still way, way harder and more expensive than email spam. (Can be swiftly dealt with, after which you have to go compromise another server)
Censorship is blocking/removing speech as judged by its content. Moderation is warning/removing participants based on their behaviour.
They're not mutually exclusive of course, but it's frustrating to see moderation projected as Big Brother, when it's actually an extremely useful and necessary too for civil discourse. Moderation can be very effective without making any decisions about speech content.
This comes down to the moderator I guess, an assumption is that they know how to distinguish the two.
The problems usually are based around where to draw the line, and the categories of content that should be moderated.
If you cuss someone out on HackerNews, you will have your comment removed. Some people call that moderation, some people call it censorship.
Also content moderation is quite literally censorship - deeming which expressions(which are actions) are appropriate and removing inappropriate.
If your domain gets added to blocklists from gmail/outlook/spamhaus you basically are.
That is the reaction now when issues with self-hosting email are brought up.
That's exactly what these big email providers should do: just make a short list of allowed email hosts, including some governments and academic institutions and large corporations, and ban everything else. What are people going to do about it? Start running their own mail hosts? Hahaha. It would solve the spam problem quickly.
A better solution would be some kind of Email 3.0 system that's designed for better security and trustworthiness, but we've already proven it's nearly impossible to get people to adopt a new open standard, so we need to stick with a locked-down corporate solution run by an oligopoly.
Bring the anti-trust sledgehammer on them?
Hahaha, that's funny! I really hope you aren't serious, because otherwise I would question your grasp of reality.
Why bother with half-measures? They should just make a short list of people to allow and get everyone else killed.
Why does this example keep coming up? Email is highly centralized these days. It is an example of how decentralized systems BECOME centralized. Not an example of how decentralized systems work. The centralization process here made email far easier to use and was critical to its mass adoption.
(centralization/decentralization is a spectrum obviously and not a binary option)
The Mastodon implementation of this is legitimately flawed.
The issue is you need users to be independent of servers.
That isn't hard to implement. Make the user ID in the protocol the user's public key and user@domain is just a human-friendly pointer. If you transition from user@a to user@b, your pointer changes -- or you can use both for an indeterminate amount of time -- but the public key stays the same and so everyone who followed you or signed up for your feed still sees it.
The current implementation is broken for multiple reasons.
First, it makes you have to care who is hosting you, because if they become adversarial or go out of business or you become large enough to want to self-host, you lose your account. That shouldn't happen.
Second, with users tied to hosts, it becomes "convenient" for hosts ban entire other hosts, which is poison because it's the dynamic that led to the bulk of email moving to GMail and Outlook. Once providers are big enough they can ban smaller ones ("for spam filtering") to exclude competitors. The protocol needs to prevent this by making the host fungible. Not doing so is a significant design flaw.
Why didn't the architects of Mastodon think of this stuff before? It should have been pretty obvious that people might want to move their accounts between servers. But now, they'll *never* be able to. The only way it can happen is if someone creates an entirely different social network like Mastodon and somehow gets everyone to switch to that instead, but it's probably too late now.
> The protocol, devised in the 1970s, can't be changed now, and we can't get everyone to upgrade to a new protocol.
This is untrue. There is a clear path to adding things like DKIM or SPF. First you add support to the major servers and publish a spec, then you give everyone else a reasonable amount of time to implement it, then you turn it on. This breaks anyone who hasn't implemented it yet, but by then that's only people who suck, and anyway so does dumping all of the mail they send to spam with nothing they can do about it.
The problem with email is that the major providers don't want to do more things like that (e.g. implement hashcash or similar), because it's better for them for the smaller providers to be destroyed than to design a protocol improvement that facilitates their existence.
Which is why it's better for it to be there from the start. But the best time to do it was yesterday and the second best time is today.
It's already too late. It should have been fixed before Twitter nose-dived. Coming up with a replacement at this point isn't going to work: the Mastodon name is already out there, and this is the time when an alternative has to be ready and usable: it's first-mover advantage.
>There is a clear path to adding things like DKIM or SPF. First you...
So why hasn't this been done yet? We've had decades.
>The problem with email is that the major providers don't want to do more things like that
Exactly my point. Email *can't* be fixed at this point because of first-mover advantage. Incumbents don't want to fix it. Users moved to email because it was available when they were ready, and it was the best alternative at the time.
Mastodon is likely in the same place now. Something better doesn't exist yet, and doesn't have any name recognition. Mastodon itself can't be fixed because the designers like it how it is: if they wanted a better design, they would have designed it that way from the outset. So if Mastodon catches on, we're stuck with it as-is.
There's nothing existential wrong with Twitter. Elon Musk generates drama which the media plays up for clicks. It's an open question how many of the people creating Mastodon accounts because a reporter mentioned it are going to be actively using them in six months. And if it continues to be the same kind of garbage fire subject to the whims of oligarchs as the networks people are fleeing, that's not likely to be many. Fixing this kind of problem is how you get people to stay.
> Coming up with a replacement at this point isn't going to work: the Mastodon name is already out there, and this is the time when an alternative has to be ready and usable: it's first-mover advantage.
Twitter has first-mover advantage. Mastodon is still a tiny thing.
> So why hasn't this been done yet? We've had decades.
DKIM and SPF were done. They're widely supported and work as designed.
> Incumbents don't want to fix it.
This is an obstacle, not a death ray. Find a way to exert pressure on the incumbents to fix the problem. Isn't this supposed to be the benefits of centralization I keep hearing about? There are only a couple of parties you have to convince in practice to get something changed?
> Mastodon is likely in the same place now. Something better doesn't exist yet, and doesn't have any name recognition.
There will be a fresh omnishambles causing people to seek alternatives a year from now after you've implemented a proper protocol for people to switch to. Do it now so you're ready for then.
> Mastodon itself can't be fixed because the designers like it how it is: if they wanted a better design, they would have designed it that way from the outset.
That doesn't follow. People frequently make mistakes and regret them later.
> So if Mastodon catches on, we're stuck with it as-is.
So don't use it until they fix it. It's not like Twitter was actually incinerated. What good is switching to a known-broken technology? Wait for the one that works, and in the meantime build it.
Elon is removing all the moderation, so it's going to become infested with trolls and bots quickly. It'll be just like email, but without the spam filtering. On top of that, advertisers are pulling out, so it's going to run out of money and implode. The future of this company does not look good.
>Twitter has first-mover advantage. Mastodon is still a tiny thing.
Go back in time over a decade and you can say the same thing about Digg and Reddit. Where is Digg now?
>DKIM and SPF were done. They're widely supported and work as designed.
No one actually uses them or requires them for email.
>Isn't this supposed to be the benefits of centralization I keep hearing about? There are only a couple of parties you have to convince in practice to get something changed?
No, not really. With centralized services, you get stuff like MySpace and Digg. Then, when things fall apart, they implode, and everyone leaves to alternatives like Facebook and Reddit. No one actually fixes problems; it's just like empires: they rise, enjoy supremacy for a while, then fall apart. No one fixes the problems they have, they just get worse and worse and eventually implode, and later something else rises up in its place.
>What good is switching to a known-broken technology?
Don't ask me, ask the Twitter exodus. I don't use Twitter, just like I was never a Digg or MySpace user. The people who value this stuff aren't going to think the way you do. They'll all flock to some new hot thing, which right now appears to be Mastodon (though it's too early to tell if it'll really catch on).
>and in the meantime build it.
Who? Who's going to build it? Don't look at me; I don't really care about this stuff and am certainly not going to invest my time in it. It seems unlikely that some company is going to build a federated social network to replace Mastodon (which volunteers already built, over the course of many years), though they might make a Twitter clone.
I will return to this comment before a head of lettuce rots.
Which makes sense to me, if you just use a system it can be hard to tell for sure if it is centralised or not, which makes it an implementation detail not a feature. Though there are definitely features that are easier to guarantee in a decentralised system, as well as features that are easier to guarantee in a centralised system, so it makes sense to look at it as a trade-off.
Sure, you can host your own email server. Doesn't mean your emails are going to actually reach anyone (i.e., GMail, Outlook, etc. email servers). There's not much point in self-hosting your Mastodon service if it's just you and a few of your buddies (might as well just have a private IRC channel).
But the website, telephone analogies stand.
That would imply that you did something so gross that everybody blacklisted your server. Currently there's only a couple instances that are blocked everywhere, and those were used solely for spam or mass harassment I think – single user instances are fine.
You must have missed the Kiwifarms controversy.
Email being store-and-forward and therefore high-latency makes decentralization work reasonably well for email.
> You can't get banned from email. You can't get banned from having your own website. You can't get banned from the telephone system. If you don't own your own data, you are at the mercy of those who do.
Sure you can. Your site/IPs can get RBLed. Hosting sites can refuse to do business with you. Network providers can refuse to do business with you. DNS registrars can refuse to do business with you. Search engines can deprioritize your site. These things happen less with personal sites than they do with mass social media accounts, but they happen, and eventually they will happen a lot more, and they'll happen to Mastodon instances, I'm sure.
Mastodon is trying to support, like Twitter and most of social media, the one to many use case, which is incredibly hard to make distributed.
Calling email a good example is like saying SMS is a good example because you can join any carrier and message any number. It's a completely different use case.
Sure there is? You can add as many recipients as you want.
Also the author is flat out incorrect on this:
> This decentralization however came with a lot of challenges and today decentralized package hosting is no longer supported by the Python ecosystem.
pip has supported github urls for a long time.
edit Further:
> The second thing that became apparent over time was also that decentralized services came with a lot of security risks. Every one of those hosts allowed the re-publishing of already existing packages. Domains that lapsed could be re-registered by other people and new packages could be placed there.
Linux package management systems solved the decentralization "problem" years ago. It's why there's so many mirrors available when you download packages. Signed packages, and even "trusted-source" checksums can provide for integrity in cases where decentralization exists.
That some package managers threw this away in favour of monolithic repositories is kind of irrelevant to how ActivityPub works.
> pip has supported github urls for a long time.
setuptools used to be able to pull packages published to the index from external URLs. Support for this was removed many years ago. Also you cannot publish packages to PyPI that reference dependencies on GitHub.
About the rest of your points on package indexes I believe I addressed them in the post already.
It's hardly the entire ecosystem, which is my point. Decentralization is there already and well-supported in the provided tools. You just can't abuse the official PyPI sources, and that's a reasonable approach.
It is, de facto, the entire ecosystem. Everything's on PyPI, and if your package can't interoperate with PyPI then it will wither away very quickly.
I’ll give an example. Rdkit[0] went for years without being able to be released via pypi and could only be installed easily via conda[1]. Despite this, it thrived enough as a project that users got its postgresql extension to be part of AWS Aurora[2] in 2020, which means wide adoption before rdkit had managed to figured out how to be distributed by pypi in 2021.
There are open source projects like this all over the place, thriving outside the centralized repos.
[1] https://stackoverflow.com/questions/58199504/how-to-install-...
[2] https://aws.amazon.com/about-aws/whats-new/2020/09/amazon-au...
What I found too challenging was having to moderate the content and finding moderators that could be trusted to remove illegal content in a timely manor. Worse, there were trolls that would use bots to post highly illegal material and then automatically submit their own posts to my registrars, server providers and government. The bots somehow even grabbed screenshots right after they posted content. I say bots because there was no way a human to perform their actions so quickly. This was a losing battle and I did not have the legal resources to deal with it, nor the development resources to play the cat and mouse arms race 24/7. I do have my own conspiracy theories as to who these bot owners were but that doesn't matter any more. Nowadays I could probably block more of those bots with techniques I have learned but I just do not have the desire to get back into that quagmire.
I suspect some of the Mastodon admins will learn this lesson with time. They, like me, will probably start in a state of denial and dismiss the risk until it gets real. And it certainly gets real.
The only technical work around I could find was to set forums to make all posts moderator-approved, meaning only the poster can see their post until a moderator approves it. This does not scale and people want their posts to be instantly available. With IRC I had to constantly add new file sharing domains to word filters to block the links to illegal material and that was also a losing battle.
[Edit] BeefWellington brings up a good point. I should add that I am referring to public instances of forums and IRC servers that anyone may join. Private servers are at much lower risk assuming the trusted members are good at setting strong passwords and static content is not accessible at all without an account and Mastodon servers are not linked to lesser trusted or non-private instances.
I doubt it. A lot of the new instances are invitation-only, and the point of Federation is I can just run my own instance and seek out the content I desire. I don't have to let anyone else onto my instance.
I should clarify that I was referring to forums and IRC servers that anyone could join. The Mastodon model in this case would be public instances that are not strictly private and are linked to other instances. Private instances would be much safer. The risk of linked instances would map to the weakest link.
But people WANT the chance of winning the "lottery" as it were, and going viral.
You're not doing that in your small discord or private mastodon.
Google Wave and Googl+ also had invite system, it didn't work out well. Gmail is exception to the rule I'd say.
Mastodon is quicksand. Instances are not guaranteed to keep existing, they depend entirely on a citizen running it, scaling it, and paying for all that, often with the help of donations. Even a relatively small influx of new users may pose an existential financial threat for the instance, or a lock so that nobody new can join.
Even when things are running "normally", instance owners may simply quit. Further, did you know that almost every instance regularly purges all media attached to toots?
Your instance, content within it, account you created within it, the media attached to your toots, are all incredibly fragile and can disappear at any time. The main Mastodon scaling approach: creating lots of small instances, makes this problem worse, not better. You're just spreading fragility.
You don't have this insecurity at Twitter, not at this fundamental level. Twitter pretty much auto-scales and your content is not lost. Sure, people may get banned, but a normy user would normally not face these existential issues.
The other thing that scales poorly is moderation. Twitter is sometimes perceived as being inconsistent in its rules (or biased), but this is a 100 times worse on Mastodon. Every instance has their own arbitrary and ever-changing rules. The same applies to federation. Instance mods regularly block federation with other instances based on arbitrary decisions.
Technically, the UX is inconsistent. One instance may work well whilst the other has page loads into the 10 seconds, or federated content delayed by hours.
Doesn't scale, fragile, and deeply inconsistent. It has a place and I still consider Mastodon an accomplishment. But it should not be compared with any of the centralized services.
I don't believe any large service is more stable in the long term than a small one. To put it morbidly, all will be dust eventually. At least mastodon has account migration baked in.
Additionally, why would I want all my content to last forever? If I want something to last forever I'll carve it on a rock, not toot it.
Pretty much everything else you listed is a positive.
Large central services give reasonable stability as to not easily shut down altogether, having your account and all content removed. On Mastodon, this is a constant threat.
I'd be interested to see the halflife of centralized services generally to be able to tease out the patterns in a way less vulnerable to survivorship bias. I recognize that invoking Google's product shutdowns is like the free space on the HN comment bingo card, but... You could also point to the cases where Google shuts down someone's entire business or life on the internet for opaque reasons, without the user having any meaningful recourse, as a "constant threat" of a system that's scaled beyond the human.
This conversation was just very, very different before the move feature. When mastodon.technology announced it needed to shut down (https://ashfurrow.com/blog/mastodon-technology-shutdown/), yes, the users learned they would lose their little chunk of namespace – old URLs would break, old content would need to be downloaded to archive – but the crucial thing is the social graph, and the move feature lets you pick up and move your followers. This is like if I could reach into the contacts files of everyone who had my email and update it. This means that "having [my] account and all content removed" is just a lot less big of a deal than when e.g. Facebook nukes someone's account.
And the thing these winners have in common is that they are enormous. They require huge revenue streams (from ads) to be able to afford the enormous investments in scaling, moderation, compliance, etc.
The idea that a billion of us can keep dumping fresh content into our account for free and that none of this content seems to be ever lost, is honestly quite bizarre even if we take it for granted.
But that's the service level billions of people are used to. And it can only be delivered by these giants.
Perhaps that's for the best? Perhaps social media in the microblogging style should be considered ephemera.
Moderation and performance you're 100% dead on.
Look at all the complaints regarding onboarding. They are forced to pick a server and many directly stop at that, loudly complaining. Two days to wait for an activation email? Unacceptable.
As for content preservation, I get what you're saying. At the same time, I do believe that people are not used to the idea that their old media is purposefully deleted, automatically. Or that their toots and instance account can disappear at will. Those are new concepts and frictions when coming from a central platform.
To illustrate that point, Facebook has multiple giant data centers that host nothing but stale content. That pic you posted 7 years ago, it's over there. You will absolutely never ever look at it again, but in case you do...it's there. These massive investments underpin my idea that users expect their content to be persisted. If this wasn't important, Facebook would not do this.
I think the complexity of content preservation is that social media has various uses. You can chit-chat and use it as a casual conversation tool, yet others use it as a broadcast channel. Some may use it for record-keeping. Creators may build up a portfolio of rich media posts. Some people may be fine with the chit-chat deleted, but for many other uses it would be not acceptable at all.
Though I can't help but wondering if this is a take by an older person or least an older way of thinking about things. Stuff like snapchat, or the stories feature is closer to what a lot of people want. Most people I'm connected with on Instagram (friends and family only) don't post anything besides stories. There are very few who want to leave a trace on the web anymore. Especially now that digging up old tweets has become a sport.
Loss and death are part of life, they should be part of online life, too. Especially for things that are basically a glorified chat room.
And for the record, Mastodon content is stored on the Great Servers in the Cloud. Most instances run on AWS VPS.
Most people are surprised that old things they did on the internet can be found, and they're perhaps right.
And to add insult to injury, better anti-harassment features are near impossible to implement because of the nature of federation. Other instances may run an older version, or simply chose to ignore your new rules.
If and when Mastodon grows to the size as becoming interesting for many mass scale harassment attacks, the only logical way out is to disable the federation part more and more. Which means you basically end up with tiny centralized bubbles. For which you could have simply created a subreddit or Discord, for free.
Moderation does scale poorly, which is why twitter is a shitshow and smaller forums (like this one) have a chance of being well moderated.
I wasn't commenting on the quality of moderation, instead on the scalability. Instance owners are easily overwhelmed and this problem gets worse as things grow.
Second, if your point is to create a tightly-moderated instance bubble, then Mastodon's federation features will largely go unused. If that is the outcome, then there's no point to Mastodon. You can recreate such spaces everywhere, for free.
You realize of course that this is true of any platform? You do have this kind of insecurity with Twitter as we are seeing in real time. Musk has already stated that he wants to reduce Twitter's infrastructure costs, he could decide that all content older than N years is being discarded, even more easily than he discarded half of Twitter's workforce.
On FB, I can go back 10 years and my content is still there. All of it. With the exception of a suspension, FB doesn't delete user content. Ever. The same applies to Twitter. Their track record of content persistence is excellent.
On Mastodon, media gets purged pretty much by policy. So you can be sure that you lose it all, thus rendering old toots useless. Add to this that every instance can disappear at any time, so besides your content, your account is also not safe.
What Musk could and could not decide is speculation. Track record looks back, not forward.
Musk has just replaced the entire leadership of Twitter. What kind of track record Twitter used to have is entirely irrelevant for how it will behave in the future.
It's the same fallacy like a reputable company selling a well-known brand. People will stick to the brand (for a while) because they associate it with the performance of the original company behind it. However, that's a deception. The brand itself will tell you nothing about how the new company will behave.
What is guaranteed to keep existing in this world? certainly not the Internet....
That's an architectural problem with Mastodon, and can probably be fixed. But twitter has it solved by acting as a sort of well-established, large data bank; a place where you store your tweet data. Going to a Mastodon instance is like going to a tiny credit union, if credit unions weren't regulated in any way.
> Imagine you're a rather small server and suddenly Eli Lilly and Company joins your instance. Today they have around 140K followers on Twitter and they are a publicly traded company. First of all with an account that large, every one of their posts will cause a lot of load on your infrastructure. Secondly though, they are a very interesting target to attack.
Why would Eli Lilly join some small instance? Why wouldn't they make their own instance? They have their own email and website after all. why be lilly@someoneelse.social when they could just be their own thing?
Because it's not worth the effort. If anything, there are going to a commercial providers that run servers for companies in a bundled way where all of their clients use their servers. But more likely, they won't bother at all with this, because it's not an important part of their business.
...aaaand we're back to Twitter in all but name
Finding is "where do I go to get X". We have URLs, which assume a specific server. There are content-based keys, such as DOIs and URIs and hashes. But how do you find where the info is stored? Google? Something like DNS? Something else? This is the hardest problem. What Youtube really sells is "discovery" not streaming hosting. There are lots of streaming services, but you won't get the views.
Identity has all the usual problems. If people can create lots of identities at low cost, there will be spam and worse. No good answers there. China has this fixed, but they don't do anonymity. You need a government ID to connect to anything.
Storage is the big cost problem. Where does all this stuff go, and who pays for it? IPFS was supposed to be the distributed answer to this, and Filecoin was supposed to be the way to pay for it. That didn't work out. On the other hand, if someone wants it out there, then maybe they have to pay to store one copy.
Delivery can be distributed, but do you want to? Bittorrent was the prototype. Peertube is another peer to peer way to do it. Each video has a home server, and large numbers of people watching the same thing won't overload it because anyone watching the video also serves it. It works OK but is not as smooth an experience as YouTube. Plus it runs down your battery and runs up your bandwidth usage. Bandwidth is much cheaper in data center bulk than out at the end of a cell connection. Maybe do something like that but with ISP level caching servers, all serving each other. Sort of like Cloudflare / Akamai.
Privacy. Messages in a private chat or group should not be visible to the operator or others. This likely necessitates e2ee, but it's possible that some multi-party protocol could protect from malicious operators without going full on e2ee (difficult on web).
Integrity: you don't want rogue operators to be able to falsify content. While you can't stop any website from faking content, you can stop it from propagating by having signatures and verifying them.
Availability: you should be able to post and fetch at all times, which necessitates aggressive caching. Signatures (from above) are useful here, because you can use untrusted storage layers.
Security: how do you prevent an instance operator from stealing your account or posting on your behalf.
Abuse: no system is anywhere near perfect, but the ones that come close are extremely centralized and use fingerprinting and behavior analysis. We probably don't want that, so what else?
Note that all of these are deeply tied into the identity problem. We would really need a form of PKI for real people (but without exposing real world identity). If people actually control their private keys, all of the above is greatly simplified. But even in defi, where custody of your private keys have a HUGE direct monetary value, most people don't want/know/care to set it up. So we have huge challenges.
> That didn't work out.
Can you expand on this point? At my end, IPFS runs along just fine: My IPFS instance in particular continues to serve its content without much fuss. For Filecoin, the content hosted there is still on the network, there's just much less speculative fervor surrounding it & other networks like it.
People should leave Twitter and start blogs so that they can talk about more complex things with more nuance
I know some of these types of groups, and I don't really publicize them, because the people who would fit in will find them, no need to grow to world-wide size.
Mastodon from the start wanted to tackle the social part of the problem about the fundamental disagreements about moderation. There's a wide spectrum between people who are keen to start Internet fights, people willing to tolerate that in the name of "free speech" principle, and people who just want to have a nice place where they can hang out without randos demanding they "debate" their right to exist.
These groups will never agree what level of moderation is appropriate. Even within Mastodon communities it's controversial what the rules for federation, bans, and content warnings should be.
Twitter, Facebook and others begrudgingly ended up being arbiters in political culture wars and must en masse decide what is acceptable to say. This ends up being an absurd situation where they try to balance the amount of abuse to an advertiser-acceptable level.
Mastodon's answer is that you can moderate your community however you want, and cut off whole parts of the network you don't like. This is ridiculed as "bubbles", but if you don't agree — make your own anything-goes instance!
It doesn't work quite well. There's still a lot to work out, but I think it's a better starting point than pleading with a billion-dollar enterprise to police the content in the way you think is right.
Those of us on it during the early days were well acquainted with the Fail Wail.
But, with a lot of engineering work - and constant upkeep - it began to handle the load. Let it be so with Mastodon.
Nearly every point the author makes is subjective. That's fine, of course. But it doesn't help demonstrate that users do (or don't) want centralization.
Personally, I'd rather be moderated by my local community than by a faceless American company trying to appeal to advertisers. And if I don't like their stance I can go elsewhere.
The solution in both cases is just basic divide and conquer. Mastodon the software may not be able to scale effortlessly to the largest accounts, but that's a market opportunity. ActivityPub certainly can accommodate it.
If inbound volume is a problem, shard inbound activities, and zipper merge on lookup, likely maintaining a cache of the most recent n amount of entries, given this would mostly apply to things like the Federated feed for a large instance where "nobody" looks far back.
If outbound volume from a single account is a problem, you similarly just split the follower list into buckets and hand off delivery to a distributed set of workers for delivery; this scales just fine with the caveat that you depend on the receiving instances being able to handle the deluge. But ActivityPub allows for batched deliveries to shared inboxes [1] you can post to, to reduce the deliveries to one message per shared inbox (which would typically best case be one per instance; e.g. the "sharedInbox" property for my mastodon account is "https://mastodon.social/inbox). Assuming all current instances has a sharedinbox property (which is likely given most of them are Mastodon, and Mastodon does), currently delivering to all the ~6m Mastodon accounts requires delivering to ~1550 shared inboxes; delivering directly to 6m wouldn't have been a problem, but it's not necessary.
If handling the outbound volume from a single account arriving at your instance for delivery to a large proportion of your users is a problem, ActivityPub already has a built in solution: To find out where to deliver, you need to use WebFinger to obtain the profile urls, and get the profile urls to find the endpoints to talk to, including the inbox, which means foo@mastodon.social and bar@mastodon.social can have inboxes on different servers without even needing to do any more advanced sharding. But of course you can do the latter too and have the same inbox url mapped by your load balancer to any of a number of internal shards.
This is all stuff we had to solve literally decades ago for e-mail.
[1] https://www.w3.org/TR/activitypub/#shared-inbox-delivery
Also, Elon has now shown us all the absolute danger of centralized platforms. Email and web servers are a federated system and function just fine, so I'm disinclined to believe so hand-waving about how this "wont ever work."
Elon showed us this? Haven't people been talking about centralized vs decentralized systems for decades?
How about resiliency? Participation from community members? It's not always distrust that drives decentralization.
I think James C. Scott's book, Seeing Like a State is worth mentioning, since it discusses the failure of centralization in more ways than simply distrust. Here's a long-form essay summarizing the highlights: https://www.ribbonfarm.com/2010/07/26/a-big-little-idea-call...
This isn't really a new idea. In building architecture, Christopher Alexander spent a lifetime writing about this. You can see some of the highlights in his keynote speech to OOPSLA '96: http://www.patternlanguage.com/archive/ieee.html ... His ideas enable end users (residents) to be able to change their built environment while still maintaining a cohesive architectural design. Sadly, his groundbreaking ideas on pattern languages became watered down into HOA design regulations; software engineers did not heed what he talked about in that keynote; and Human Computing Interaction design dropped end-user-customizable software (like Hypercard) in favor of designs that favored aggregators, because it is more profitable (not necessarily more resilient, or better for society)
Citation needed, the thing seems to be scaling just fine for the needs of the current userbase.
> There also is no real agreement on if larger or smaller instance are going to make the problem of moderation better or worse.
That's the whole point. If you don't like how a community conducts itself you're free to block them as a user, or even as an admin.
> A service like Twitter is much cheaper and easier to run if it does not have to deal with federation on a technical level
We had that, it was called the internet. Everyone went to Twitter, because they want to talk to everyone else without a hundred accounts. We still have that, it's called Mastodon (or pleroma...), you don't have to federate if you don't want to.
You don't have to like it. If you have ideas, implement them. But don't tell the network to change itself because, as someone said once, they were there before you and they'll be there long after you leave.
When has any medium gone from centralized → decentralized?
https://en.wikipedia.org/wiki/Breakup_of_the_Bell_System
The companies we now know as AT&T, Verizon and CenturyLink all came from that.
jaredcwhite: When you say decentralization is inevitable, were you talking about as a result of government intervention?
Obviously some aspects can be more efficiently centralized. What matters is who controls the centralized part(s). E.g. There are still a few things to be re-learned from Usenet and Fido.
Ultimately, that's what gets me about all these sorts of conversations. People will post articles on websites saying the web is dead or whatever…same thing happened 20 years ago, same thing will happen 20 years from now.
Don’t get me wrong, I’m a huge fan of decentralization. I do systems engineering. Decentralizations means redundancy. Redundancy is good.
But if all you’re doing is federating a mono-platform/algorithm, then the single point of failures move to the platform itself. Decentralizing a uniform platform is like replicating the same human being lots to improve the redundancy of the human race. Ask the gene pool how well that works.
Variety is good. It’s what brings beauty to the human experience.
The federated feed is nothing more than a combined list of all public feeds from all the users on the instance. There is no way to get a central firehose, since there is a large number (I was going to say majority, but that's probably unlikely) of posts that never even leave the local instance.
What the author sees as problems with mastodon (decentralization, moderation differences between instances, etc) to me are the strengths of mastodon. It's the whole dang point.
Most mastodon instances aren't trying to be another twitter. The ones that are and do want to scale will inevitably run into scaling issues that any large platform full of user-generated data will run into. I'm just glad that there are different instances so when one big instance runs into scaling problems, other instances can just keep rolling along unaffected by it
I think talking about mastodon as a singular entity is fraught with problems because it's simply not that
Whether decentralization for social twitter-like sites might work is up to discussion, but Mastodon is just bad try at that.
You're trading consolidated moderation by one rule for thousand little fiefdoms, each with different rules and waging ban war on eachother. Empower the users to filter and pick what they want to watch, not moderators
That's already the case under the "moderation by one rule" paradigm.
I can imagine a friends and family instance that solves this: only invite people I wouldn't need to moderate.
Honestly, though 1 person per instance makes more sense to me. There needs to be a way to drive the (time and money) cost of hosting an instance to zero.
There probably also needs to be some sort of web of trust style reputation system. (Or just a way to whitelist users instead of blocking them.)
Frankly distributed social medai should be more akin to Google Reader and less of what Mastodon tried to be.
Like, have server have its users stream, still allow moderators to block incoming comments on that (aside from mod ego trips bad actors will always exist), but allow user to subscribe whatever they want without mod touching that path. So say mod would be allowed to block @server or user@server and that would cut that from public stream but not from people that subscribed to the guy or to feed from that server's public streams.
That way you can as mod still choose to not have politics poison your stream of dunno, amateur woodworking twoops, but it doesn't infringe on your user's freedom to see and talk with whoever they want.
> There probably also needs to be some sort of web of trust style reputation system. (Or just a way to whitelist users instead of blocking them.)
Frankly blockchain here would be the sole case where that technology might've had any benefit, even if blockchain was just used to store user's identity data and facilitate profile transfer. Maybe separate one if some servers want to make a blocklist in "if at least X out of Y servers see user misbehave, permaban" way. But eh, what stops someone from making 20 accounts on 20 different federated servers and report same post ?
Distributed reputation is hard. "just" downvote/upvote karma system is frankly silly to be anything else than brag number, as essentially you just need to make popular post to counter few unpopular ones. I think there should be a system to attach up/downvote count to a post, that makes it easy to cut on low effort/dumb content without much mod input, but making that not be gamed cross-server would be hard.... and per-server score is less and less useful the smaller server is/
I’d argue that some of the things he sees as problems are actually features.
Email as a decentralized medium has survived for decades. You use a big provider like Gmail, choose a host in another region, or an organization like Proton Mail that does thing somewhat different.
It's OK that Mastodon is messy and at times chaotic. It's organic.
As oligarchs run Twitter and Facebook, decentralization seems less like a goal, and more like a necessary condition.
> Imagine you're a rather small server and suddenly Eli Lilly and Company joins your instance.
Why would they ever join a small indie mastodon server? Do they send email from elililly@hotmail.com too? If they can run an email server for @Lilly.com, they can run a mastodon server for @Lilly.com
Decentralisation is not, as this piece states, 'a Questionable Goal'. It is the essence of a thriving internet, it is what makes the difference between a world-wide inTRAnet and an inTERnet. It makes it possible to pick and choose your own services or host them yourself while still being able to interoperate - the 'inter' in internet - with others. It may not be a good fit for someone's business plan but to that I can but sing a song while playing the world's smallest violin: Nae kings! Nae quins! Nae lairds! Nae masters! We willna' be fooled again!
Because Mastodon does not, today, solve any of social media's problems
This is an unbelievably generic straw man.
People have mentioned email, but what about DNS? RIPE/ip4 routing?
Small, federateted invite-only, self-hostable communities seem exactly the right fit for social media.
The gatekeepers know their users (invite only).
Instances can be small, easier to agree on a culture/moderation policy.
Federation allows cross-pollination of ideas across communities.
For example, with DNS most people (with internet access, I think) can register a domain, but not everyone can register a .no/Norwegian domain. The organization (NORID) that has delegated authority for the zone, knows Norway. It has a priv.no subdomain, and it knows about government entities an municipalities that we've decided should have special status. This is a pretty great example of efficient decentralized authority.
There are other registrars that have decided to organize things differently, like Tuvalu - but that doesn't really impact NORID as such.
Start with the principle that you own (and pay for, if necessary) the cloud resources for your piece of the social network. A standardized API that does the things you expect of a microblog; post stuff, collect replies, followers, etc. runs in your account. You control it: access, moderation (possibly delegated as you prefer), advertising, throttling, whatever.
Assuming everyone is using a standardized protocol (DIDs for identity for instance) users could interact seamlessly. How hard could it be to clone Twitter on Cloudflare? Would it take more than a few thousand lines of Javascript/WASM to replicate the basic microblogging functions? A good onboarding system to automate the setup to be at least as easy(...?) as typical social networks would be necessary. Search would solve itself if the network was sufficiently popular.
In none of these cases is the debate resolved by asking: "What problem are we trying to solve?"
(edited for clarity)
Email is similar and now most people get their e-mail from one of a few providers who exploit their data for advertising purposes. This is exactly the problem we wish to distance ourselves from in regard to centralized social services such as Facebook, Twitter, Youtube, etc.
If you're not the customer, you become the product.
Mastodon and the Fediverse made decentralized software that people want to use - but they didn't come up with a way to make it sustainable. Web3 has a model of financial sustainability, but they've yet to come up with any other decentralized software that lots of people want to use. These two communities focused on decentralization should really collaborate more.
I would happily pay a bit for Mastodon if I needed, but for me Twitter is good enough though for now.
I would argue that in fact, cryptocurrencies famously face similar challenges to the fediverse with the moderation a decentralized community. Fraud, abuse and grift are rampant in the cryptocurrency community and even the good actors in that space seem to struggle to find any way to reduce that. The only limited successes there are from the centralized exchanges.
Bitcoin has brought out the worst in many people, to the point where coal power plants were coming back online in order to handle the power demand of generating bitcoin, and miners were grabbing up as many video cards as they could, boxing out gamers from reasonable build budgets for years. Bitcoin is how society runs when it only cares about financial incentives. There are plenty of parts of society that run without a financial incentive at all just fine, and I'm hopeful the future of the web is a lot more like Mastodon than it is like web3.
In as much as any vague buzzword can have meaning, Web3's chief innovation seems to be creating a model of software that actively repulses a whole lot of people, e.g. play-to-earn video games where the financial incentive turned a fun and enjoyable leisure activity enjoyed by millions (i.e. video games) into a combination of a joyless job that pays less-than-minimum wage, digital feudalism, and a pyramid scheme.
These kinds of financial incentives have a habit of driving out non-financial incentives, and the latter tend to be what makes online communities any good, so it shouldn't surprise you to learn there's lots of animosity to web3 stuff in a lot of mastodon/fediverse-type communities.
The short short version is that it scales just fine if you use it the way its intended. The protocol provably scales just fine (contrary to his claim), the real problem with scaling is that it violates the core idea that Fediverse server apps (all of them) are written with the idea of serving a community, not hosting the next multi-millon user twitter competitor. There's more to it than that, and there are a lot of straw-man arguments and contradictory claims in that article.
> For that particular problem the solution might be one of those two things, or something in the middle. But out of principle it should be neither of those two things.
So which one is it? Pragmatic implementation detail or "out of principle"?
Also, if the author prefers a solution somewhere in the middle, wouldn't the Fediverse be the perfect example? Full decentralisation would be some kind of p2p network or possibly blockchain thing where there are no servers at all. Instead, each server on its own is perfectly centralised, it's just the community at large that is decentralised.
Its interesting - most recent examples this is true, but it seemed to work well in the past. Bit torrent is a staggering success. Email is showing some flaws now a days, but how many other application layer protocols do you know that were designed 40 years and still widely in use.
Although controversial, i would actually say original bitcoin was also a success in the sense it did what it intended to do beyond anyone's wildest dreams (whether or not that is a good thing is a different conversation)
Decentralized protocols seem to have a long history of success. Somewhere that seems to have flipped.
There is no way to inject spam and ads to the torrent and little incentive to fake seeders/peers to make people download the wrong one. That's why. It's not a communication method, it uses other communication methods to distribute torrents/magnet links
> Email is showing some flaws now a days, but how many other application layer protocols do you know that were designed 40 years and still widely in use.
Frankly we nearly had that in blogosphere. You could federate what you want to watch via RSS. You could discuss under the post. There was even mechanism to get which post referred yours via pingbacks (but of course got removed due to of course being way to distribute spam. The thing it lacked is some kind of place to be endlessly fed whatever is now popular (as it is trend for every popular social platform) and general seamless usability.
But it put entirety of curation up to the user, don't like the blog ? Don't subscribe it, no mod to tell you what to do nor someone injecting shit into your stream because mods decided this thing need to be promoted now
What people are after in social networks increasingly is local communities. Whether it runs on protocols or platforms doesn't really matter. I'd agree though with the authors and Moxie's take that protocols have a lot of problems due to their slowness and technical hurdles. In reality they also tend to coalesce into platforms anyway due to economic pressures.
The most appealing quadrant for social networks I think is the Reddit one. Some platform with a baseline and minimum of largely legal rules but local moderation. Which is essentially how real world federations work anyway. If it's a new or existing one I don't know but I could well imagine that Twitter or Facebook more and more devolve power towards regional communities and create better and fine-tuned moderation tools. Telegram and Discord are already examples of this as well.
While I don’t agree that decentralization itself is a dubious goal, the argument that moderation, the legal side of things, etc… is not scalable (moderation especially in its current form where the entire fediverse can report anything) seems quite true? We can’t rely on generous people hosting instances, taking legal risks and spending time for it if the fediverse wants to grow bigger than what it currently is.
One seemed to be that different moderation standards make a coherent experience impossible. I think this is a fundamental difference of the fediverse. There is no one experience, and that's a feature not a bug.
I think it also makes out that by banning other instances outright, instances will be largely disconnected. I think this ignores that, more or less, people agree with what is acceptable and what is not. And if they don't largely agree, maybe those groups of instances should be disconnected.
I think it also assumes that this disconnected experience is somehow opaque to users, when I think it's pretty easy to figure out that you're only seeing Toots from your own server. I think that's also not a bad thing. Again, if you set aside the idea that everyone has to be on the same experience, viewing the same content, the decision calculus becomes easy. Are you happy with the depth and breadth of the content you're seeing? No? Then move instances (which is easy).
My other criticism is about legality. The US has section 230 of the telecoms act, which basically allows internet companies to moderate their websites, without having to be responsible for what is posted there by users. Basically, the argument that the law will somehow sink the fediverse ignores that we made it this far. The web was allowed to flourish, in part because what you say online is largely your responsibility, not the responsibility of the sites that host the content (with a few exceptions). I'm not sure how Mastodon meaningfully changes that.
Maybe Lemmy [1] with a Twitter UI is worth exploring?
The financial system has a lot of experience with dependency on centralized organizations that are too big to fail. It's trading one set of problems for another. In particular the risk is that the organization starts to "fail" but not badly enough to cause a mass collective shift away. Things just degrade and become terrible but there's never a moment that overcomes the enormous activation energy needed to migrate away. With a federated or decentralized system it's easier to bleed off from an institution or service that's started failing at its core mission.
Given the history of email, I'm surprised there hasn't been more about the economics of spam, blacklists, and so on.
Substitute the email-specific parts of the following article with Mastodon-specific terms and concepts, and it doesn't paint a very pretty picture.
Off course such concepts of less independent instances are a different mindset.
Any fully distributed system of independent instances must end up in some "maximization" of bandwidth usage.
@doener also, if you think about such cache servers (editing content will be a 'problem'), one should optimise the then smaller messages interchanged between instances and between instances->users in a way, that these fit into the least count of Ethernet-frames. In an optimal world: one!
Maybe in a BitTorrent way, users could also act as "cache servers" for content."
But for old heads, I think the real goal is for something more like the blogosphere or usenet groups or forum cultures where it wasn't about lack of trust but the ecosystem felt more dynamic because there were simply a lot of people just kind of doing their own thing in different ways. It iterated better, and it iterated in ways that were more centered on the content creators and consumers instead of a platform clearinghouse.
The articles point seems to have its flaws and maybe the main point is that mastodon will be ruined not saved from ruin by the scaling problems being outlined though so :shrug-emoji: I just wanted to note that initial reaction / feeling.
This is an inherent problem that centralized platforms try desperately to handwave away. However, sometimes handwaving doesn't work, and there's no lowest common denominator to fall back on which will reliably keep you out of trouble everywhere; for an example that probably won't get me flamed too badly here in this decade, is it Derry or Londonderry? Just calling it Stroke City won't work if the people on the site are sufficiently enthused about their discussion forum being Correct. It gets worse the closer you get to contested claims of genocide (do you ban the people claiming it was genocide or the people insisting it wasn't?) and discussions about who gets to qualify as human. Trying to be apolitical only works if you ban all the people from one side or the other.
It gets worse once you globalize, and you run into governments that are really, really keen on enforcing Correctness on certain issues, especially if you want to expand into two countries with mutually-incompatible views of which side of a given issue is Correct. Silence is not always an option. At that point, decentralization is the only solution.
There are pros and cons either way, but finally it seems that we have a choice. The Fediverse is here and will remain, and seems to have enough gatekeeping to limit federation of unwanted/spammy/malicious servers. Centralised forms of social media will continue to exist but this new landscape will change how they work and will change the content created and shared.
This folks is all you need to read from the above article.
Yes, I'm exaggerating a little bit, but we keep forgetting that federated systems have worked before and will continue to work. This is not some failed idea.
Email.
> a better alternative...would be...an Open Source implementation of Twitter...that being Open Source would...[let] different communities exist side by side if we can't agree on common rules.
Isn't that literally Mastodon? If you don't think mastodon's tech will scale or you don't like how the protocol is developed, that's one thing. But I don't understand saying that you're anti-decentralization, and then proposing decentralization as the alternative.
There will probably be work in that domain sooner or later. Or it’ll never really take off. Either way I like how it’s small and niche. Let the normals stay on Twitter and they can keep the advertisers, too.
Moderation is done by users of a node or a server admin or admins. And from what I can see admins block on a node level and users can block folks at the user or a whole subdomain/node.
From there people can choose to run external services to further handle load, but right now just to launch a small instance, needing all the crap that comes along with the node.js world sucks.
The blog post isn't really about how to technically scale a service though.
I still remember having to run his malady of a web server when we still had big Ruby team. I guess it might've been "good" compared to other things Ruby...
and the clown car mentality like
> had this stupid setup where the people deploying Rails didn’t have root access. I told Kevin that this was stupid and apparently it was Kevin’s idea because he told me “fuck you”.
Like, holy shit, even at dark times of cgi-bin we didn't had much of this idiocy and he's taking issue with people not wanting to deploy stuff as root...
The author should take a look at Linux Torvalds talk with Google (2009). https://www.youtube.com/watch?v=4XpnKHJAok8 back and day when development was centralized (no cloning).
The concept of "Trusted Circles" is at the core any resilient information exchange and it is by definition: Decentralized.
Until we decentralize those we're going to have problems:
The second half made me think, though. Specifically, about the difficulties around back pressure and moderation.
The final thoughts are on the right track. A central foundation that acts as a steward for a protocol or platform, and instances of that protocol/platform are decentralized and federated.
Done: it's called the W3C. https://www.w3.org/TR/activitypub/
> or platform
Done: it's called Mastodon gGmbh. https://blog.joinmastodon.org/2021/08/mastodon-now-a-non-pro...
This is perhaps the greatest weakness of mastodon. On one hand, there need to be some mechanism to disable harrassment, on the otherhand, it's practically impossible on "decentralized" networks.
This isn't framing the problem the fediverse is solving accurately. What fedi solves isn't not trusting anyone, because it's not a fully P2P network. Rather, it's more about having accountability for the services you use. With a massive centralized social media platform like Twitter where it needs to effectively scale up to be able to accommodate the entire world, it's impossible for a service like this to exist without either being run by a government or a corporation. In the US, the two are more or less the same anyways since we've outsourced most of our infrastructure to the private sector and have been doing that for decades. And in the case of a private entity controlling one of these services that need to scale massively, it's effectively impossible for them to not be incentivized to do all the things that make people not trust them -- mining users' data, etc.
On fedi, you know who your server admins are; they're human beings that you can message if you have a dispute with another user, or there's an issue with the service itself, or whatever. If an instance admin does something shitty, they can be held accountable for it and can have their reputation ruined on fedi, which isn't a perfect system since it leads to things like people being defederated unjustly, but I would say it's far more scalable socially to have a patchwork of different small to medium sized servers run by individuals or teams of people rather than faceless corporations that largely automate all of their user-facing interactions (algorithmic bans and support) where it's basically impossible to hold them accountable for anything unless you have millions of dollars to burn in court.
It is true that centralization/decentralization are implementation details for solving a problem, but this is a case of technological solutions not being sufficient for cultural problems and vice versa. Fedi, for all the problems I have with it, is however closer to striking a balance between solving problems on both of these fronts because it returns the internet to being a community-driven network that has its own culture, and has more ability to handle bad actors or failures in the network with federation as opposed to earlier eras in the internet where for example everyone was using their own separate forums and the centralization of the internet onto social media platforms was really an inevitable consequence of that structure of the internet.
Dystopian, actually. It promotes (and depends on) a nihilistic view that trust cannot be built, that it can only decline over time. This may be true or false. But it's not 'utopian'.
(What was that segue about python package management intended to show?)
considering how hard it is to scale centralized such systems, it makes simple sense that it be much harder to scale a decentralized system.
That's not to say ActivityPub isn't a pretty trash protocol, insecure by default (no E2EE), wide open to abuse, and very wasteful (because posts have to be replicated in every server, making unnecessary copies of a post that aren't then distributed via P2P).
But the social issues of pretty big Mastodon instances blocking other smaller instances with complete impunity, means the Fediverse as a community (not a protocol) encourages bullying and isolation. I find this extremely troublesome in the wake of the Capitol attack, fake news campaigns, State-sponsored disinformation campaigns and everything else that is wrong with social media today.
You will get your server banned from other servers if your server doesn't ban the servers these servers ban (second-grade sanctions). You can get banned for not moderating enough, or for incompatible code of conducts. You can get banned for an instance that is too small, as other admins believe you are using it for ban evasion.
The worst part is the user can't know if the posts they make will be seen in other servers. They can't know if sending messages to other user will work. They can't opt out of these admin-imposed forceful defederations.
And even worse: it is absurdly simple to push illegal content like child pornography into other servers, making these servers store illegal content in their hard disk. And the only way to avoid this is to have server administrators manually see the offending post and possibly block the offending domain. Perhaps with the help of users reporting this post.
If Mastodon is the best Twitter alternative, I think nothing can replace Twitter as it is today.
Mastodon doesn't need to scale like Twitter, the point of Mastodon is many instances, not a single giant one. Sounds like Mastodon is not for you. Good, you can leave now. Mastodon's mission is not, and never has been, replacing Twitter.
It'll serve a niche group then be put out to pasture.
Isn’t this exactly what Jackattack is working on right now?
Mastodon and ActivityPub have been flying under the radar for years and now this is its close-up.
The ActivityPub spec hasn’t been updated since it became a W3C standard 4 years ago; no doubt it needs to be updated and probably will be, now that so many users are now on ActivityPub servers.
There are going to be growing pains but that’s to be expected. People remember the days of the Fail Whale, when during a big news event, Twitter would just stop working.
Timing is important—there’s enough visceral dislike for Elon Musk and users getting tired of billionaires upending things.
It’s not going to happen overnight but I expect good things to continue to happen for distributed social media.
When it comes to scaling in particular, we have a counter-example in e-mail which shows scaling an inbox model + reflection (mailing-lists) to a follower-list can scale to a vastly larger audience. Of course it can, given that it decomposes neatly into a trivially parallelisable set of components of a well understood nature.
It was when Twitter was still getting their fail whales, sometime around 2007, I first pointed out that parallel (might even have been on HN). There are some tricks to avoiding lag in presentation for accounts with huge numbers of followers, but there decentralisation already serves to turn a flat list of followers into a tree of instances -> subsets of followers per instance, and if/when any instances gets to a size when scaling a single instance becomes problematic, then decomposing into a virtual federated set of instances presenting a unified namespace solves that just fine (have built a mail setup doing just that - mapping <user>@<domain> to <user>@<backend shard> is not hard)
The issue of "unpaid Labour" and peoples lack of desire to self host also depends on an assumption that Mastodon is inherently tied to unpaid volunteers or self hosting, but you can already pay for Mastodon hosting, and more options are sure to appear if it keeps growing. And the largest Mastodon instance is run by a non-profit getting paid by sponsorships already.
And this goes straight to the issue of what happens when someone popular joins an instances and/or trust: No instance needs to accept someone who brings undue load without expecting donations or charging or just turning them away, but the converse is that trust is layered in Mastodon:
You can trust someone on the basis of a history of behaviour and integrity of the node you're on, but people who need more trust can also run their own instance on their trusted domain. E.g. if POTUS wants a trusted Mastodon account, the logical choice would be to have webfinger for @potus@whitehouse.gov or similar point to a trusted instance run by the government. Nobody else can set up addresses on whitehouse.gov because it requires controlling the ability to run a webfinger setup responding at https://whitehouse.gov/.well-known/webfinger?resource=acct%3...
Conversely, if people need someone trusted to run their instance, there are any number of people who'll happily take their money to provide guarantees.
I am sad that the crypto/defi/web3 crowd hijacked the term "decentralized" and now people who should know better equate the two.
Web is decentralized. E-mail is decentralized. The internet is decentralized.
> Mastodon encourages not just decentralization, but federation. [...] I will make the point that this is the root of the issue here.
Ok, so we're actually talking about federation being bad, not decentralization as such.
> I used to host a pastebin for a few years. It was Open Source and with that others also hosted it. I had to shut it down after it became (by a small percentage of users) used to host illegal content. [...] I really hard a hard few weeks when I first discovered what my software ended up being used for.
We take that risk every time we open source something or run an internet service. I would not be surprised if some criminals somewhere used Flask for nefarious purposes. That doesn't mean Flask and his other software is not, on net, a huge benfit for society. This issue is orthogonal with (de)centralization, tho.
> Imagine you're a rather small server and suddenly Eli Lilly and Company joins your instance.
Let me rephrase that to show what a straw man this is: Imagine you're a rather small email server and suddenly Eli Lilly and Company starts hosting their mail on your instance
The chance of that happening is precisely zero. WTH would they do that (and why would you let'em)? The same argument goes for Mastodon.
Turning to the title of the post, Scaling Mastodon is Impossible, I agree with the author (to the extent that I know about Mastodon, anyways). Maybe Mastodon isn't up to the task, maybe ActivityPub as a protocol is inadequate, but those are technical challenges to overcome. I don't believe this dooms the entire "decentralized federated network" concept.
> Wikipedia for all it's faults shows quite well that a centralized thing can exist with the right model behind it. [...] A “Not Twitter Foundation” that runs an installation of an Open Source implementation of a scalable micro blogging platform
This is an interesting proposal. I worry that the cost structure - in terms of hard work, not server capacity - is way different. For all its edit wars, the basic principle on Wikipedia that someone (or a group) authors an article and thousands or millions people read it. Once the article is written, it is (for the most part) static. "NotTwitter" is the opposite - you have constant stream of new content that needs to be policed (if you don't want your town square to descend into madness). Reddit's army of moderators shows how difficult that task can be.
[Meta: was the article flagged? It's got a fair number of points but is way down in the list of articles]
The analogy to decentralized package hosting seems misaligned. If a package repo disappears, it breaks a tree of dependencies. If a social media instance disappears, the followers (the "dependencies") can adapt without being disrupted. This seems like a feature.
On the issue of "not agreeing what it should look like," this also seems like a feature. Communities can use the platform to self-organize according to their own visions and priorities. Surely, someone who has spent time exploring the various instances available can quickly identify this. Not agreeing what it should look like is what makes it what it is. Not having to force everyone to agree what it is, is what it is. Again, it's a feature, and something fediverse users have happily "learned to live with," and moreover, embraced.
Moderation and illegal content are a legitimate challenge on any platform. With the decentralized fediverse, everyone handles it according to their own policies. This does mean some admins allow it perhaps more than they should, but it also means they face the consequences directly. Admins who work to keep it out do face a daunting task in many cases, but there are tools for it. And a process of sort of natural selection already seems to favor those who have the resources and energy and motivation and skill to do it well. Seems, again, like a feature.
"Mastodon is old." Not sure how this has anything at all to do with centralization or decentralization. We all use things that are old. Many old things work great. As far as others not getting their changes merged back, well, that's the benefit of forking and open source. Anyone who wants changes that aren't part of the upstream can fork it and make those changes. If enough people like them, people will move to the fork. It happens all the time. Feature.
Technical challenges described are demonstrably surmountable.
Market-based approach: well, we're living it, and Mastodon in particular is succeeding in the market. There are competitors like Pleroma and microblog.pub. If enough people decide Mastodon proper has "fucked up too much," then there are others right behind it ready to "step in and replace it."
I dunno. There are arguable advantages to centralization, but I'm not sure this article touches on them, or at least not convincingly. Rather, it seems like the main complaints, if they're not in fact addressed by the design and operation of the fediverse, are design choices that give the fediverse advantages in other aspects. In any case, I don't see how the claim of the headline is well-supported by the points raised in the article.