The Fediverse is inefficient but that's a good trade-off
berk.es
berk.es
Makes me want to implement some ActivityPub features on my own site, so I can get in on the fun!
But I'm also seeing authors, actors, journalists, doctors, and scientists I follow make their way over, as well as a few novelty accounts.
Sure, many of them are dual-posting. But I still see them interacting directly on Mastodon and not just mirroring their Twitter posts on it.
I didn't expect this, but I'm finding it very easy not to use Twitter most of the time. The main exception is friends and family messaging various tweets to each other to share news or laugh about something. So I still have an account, it's still useful, but I'm not scrolling it casually anymore.
Are you seeing your mom and dad moving over? Or your teenage kid and her friends? Or your non-tech friends?
Teenagers and even 20-somethings are already a lost cause for Twitter. They're on TikTok, Snapchat, Instagram, etc.
And yes, I'm seeing artists, writers, and other non-tech people moving over and finding instances that better suit them.
My friends who consume Twitter as a passive way to be entertained by celebrities aren't moving yet. I'm not sitting here trying to say that Mastodon is a drop-in replacement for Twitter for everybody. But I am surprised and encouraged by how easy it has been for me and many others to find a better experience.
If having users already there was enough, Google+ would be king of the hill.
I mean to say, that having existing users is not a panacea, since Twitter literally bought Vine with all its users, and then proceeded to squander it. And no one else made the short video format work (<10 minute) for a long while until TikTok gave people a kick in the pants again.
Eternal September lead the the collapse of USENET largely because no company could control it, thus it got harder and harder to manage / control in a way the normies wanted, and governments wanted.
USENET today is primary binary transfers, i.e the high seas of the internet.
It is not largely a communications system anymore.
We are already seeing that as people leaving Twitter, the nice left censored authoritarian bubble and find out that the wider internet is not left censored authoritarian so they are demanding every growing block lists of servers attempting to make a little walled garden for Activity Pub, which undermines the entire purpose of federated/ distributed protocols
Fediverse has all of the same problems inherent in a distributed store and forward system and may suffer the same fate, but lack of centralized control is not the defining problem here.
And the underlying software has had the time to work out a lot of the nuances of this design and how to get it to a larger scale successfully. It could still break if the whole planet took up ActivityPub, but it's a hell of a lot closer to the ideal than what happened with the pre-Web protocols.
This meant efficiency of scale plus Ricardos law kicked in and Roman food production increased even without Haber or John Deere.
But most small farmers throughout the ages did not do this - they could have invented the co-op mass farm but the risks were too great - there was not a trade system globally that could guarantee delivery of food stuff X and so it was sensible to not focus your own production on food stuff Y but to grow a little of everything.
I think what we are seeing here is Mastodon etc as the response to the risks of centralised but efficient servers - privacy, ads, government monitoring etc.
But I absolutely take the point - anything that can fit in an online comment is a gross simplification- especially for "farming since invention of agriculture".
Also the 20 mile is not a hard limit - it's easy to envisage some kind of system where I grow my cash crops near transport like river or coast, and sell them on a trade network, and you grow different crops further out and bring them not to the river / coast but to me to feed me. That probably will be more efficient but just look at the network of trust and possible disasters required - only very stable politics would allow anything like that - and that stability was rare. China probably had more of that for more of the past 2000 years than anywhere else - and that may be the cause of the invention of the bureaucrat!
The Danube and the Rhine were the two major rivers that could be navigated by Romans and they were the borders of the empire for that reason: they were the only way to supply border garrisons efficiently. They also had the Nile which isn't _in_ Egypt, but _is_ Egypt.
China by comparison is three major rivers which change course frequently and violently: https://www.youtube.com/watch?v=QFsQoY32n5g
I don’t see why the southern shore of the river would be easier to keep supplied from those rivers than the northern one. So, if the empire ended there because of supply issues, wouldn’t it end a short distance north of the river?
Also, wouldn’t those rivers only help supply from other border areas, not from the much larger inner parts of the empire?
I would think those rivers just are natural boundaries because they provide some protection against invading enemies.
Retreat is easier when not drowning.
Brett says that moving grain by road would double the cost for every 100 miles. He has quite a few articles on premodern farming, and may have given a different figure in other articles.
Roman roads were high tech for the era and those were still pretty bumpy. A modern road is much more efficient. Bouncing over uneven terrain not only damages the products and veg it also causes wear and tear on the cart, the animals, and the driver.
Also doubling isn’t a bad thing for the first couple of multiples. Each doubling increases the supply by a factor of four on open terrain. It’s no accident that big cities are at the confluence of trade routes. Up in the mountains creates a funnel that raises the average cost and lowers the reliability of everything too much. It’s like you’ve blockaded yourself.
A course that goes over how everything we think is bad today is good in a world stuck in the Malthusian trap. Working harder today to have a better life tomorrow just means your children working harder tomorrow to have the same life you have today.
The option he didn't talk about at all was how in such a world sexism could keep average living standards quite high so long as all women of reproductive age were kept in near starvation so they could barely carry children to term. But there's only so much unthinkableness you can expect from someone living in the current day lecturing at a university.
Definitely sounds like a reference gold mine.
I may have misunderstood the tweet of the review of the article based on the book.
Anyway, my general conclusion is that however hard it is to understand history, understanding economic history is twice as opaque
The above quoted book tried to argue that the poor dying continued and the middle classes of victorian briton had more kids and hence took over - a sort of poorly argued eugenics wrapped in the usual "it's culture innit" argument.
However Like I say I have only read enough of the book to decide it's in my reading list only after I discover immortality.
That a simple mathematical fact like that is considered eugenics is reality denial on par with the flat earth society.
The 19th century was special because people figured out how to use coal to do the work done by muscle in the middle ages. The previous centuries had seen a continued decline in capital costs. E.g. mediaeval interest rates were routinely 15% and more for secured loans.
When I type a hashtag into the search box of any given instance, I get useless search results.
For instance if I search for #lisp on fosstodon, I get some hash tags that begin with #lisp, not #lisp itself. They all have very low activity, from several years ago.
If I do the same on another instance like mstdn.ca, I get nothing.
BUT: if I use the URL <domain>/tags/lisp, bam: recent stuff appears! On either instance. The same if I see the hash tag in a post and click on it.
Users are going to join Mastodon and try to use the search box to find content related to their interests, and falsely conclude that there is nothing, or next to nothing. That is going to cause some nonzero user attrition.
It's not just a problem for tags like #lisp, but mainstream things like #kids, #parenting, #dogs, what have you.
My home instance comes up with nothing for any of these searches, with or without hash. Not quite; for "dogs" it finds a user named "Elon Musk F___s Dogs" as its only search result.
It's not self contained and after running an instance for a while you can be in update hell because the containers might need newer versions and you need to read up on every service and how to update that.
I much more appreciate the gitea approach. Simple setup, self contained but can be extended with external databases if needed
Sounds like Mastodon is missing some engineering discipline and maybe this is a good time for some to show up. Fixing some of this stuff before Twitter gets even worse would be a good thing.
What I am now completely oblivious to is KPIs to judge the health of the components.
It’s practically the history of open-source, unfortunately. Mozilla was able to resist that for a while but is unfortunately falling into the same trap
And anyway you can search for users by substring and then follow them and harass them.
Like in my example whereby I was searching for a dog and it found the user "Musk F___ Dogs".
I think the restriction is on searching for posts by substring of content.
It’s all fun but please keep this in mind. Perhaps host your own instance?
It would feel better if we had one great feed reader, and an authoring pumping station.
The blurred lines makes it more difficult to comprehend as much.
Who moderates offshore content?
Whoever is in the pilot seat and that’s beautiful. Although if things start to get a bit too edgy we might see entire servers blacklisted from store apps. Also server-wide block done by administrators. If I’m not mistaken the largest mastodon instance is Pawoo.net and that’s basically a japanese board for posting controversial hentai artwork. Pawoo.net is blacklisted from most popular Mastodon servers.
Slightly related Google blocked Element a few years ago because something posted on a federated instance. I think they strong handed matrix.org to implement moderation on the default servers.
I believe - although have not verified lately - that Tusky will not connect to some specific instances that host degenerate people.
I use Fedilab instead.
Mastodon at least is not trying to shove politics and celebrities down my throat.
Oh yeah, that worked out great with the planet-destroying overlords that are the oil companies. No, this is a terrible heuristic.
Self hosting your own instance is great when possible (and I definitely think it should be possible for almost all services, at least in an ideal world), but you won't be able to do much in life without trusting at least someone along the way.
I don’t want to be in a position where I’m defending twitter but need to remind federated instances are run mostly by normal people. Having the cake and eating it and all that.
In terms of utility for social networking, there's no defending Mastodon relative to a Facebook or Twitter. The small userbase is reason enough not to switch; what use is a social networking site if the people you want to interact with aren't on it?
That there are so few if any people who both recommend Mastodon and don't have moral/political objections to Twitter is damning. Where are the objectors who recommend it on purely descriptive grounds? It's a bit like GNU - it's cool, but don't try to tell people it's a viable replacement for Windows.
This isn't different than other websites. It's true of web forums too. It's fine, just don't post anything you can't afford to lose.
Experience: on my third Mastodon account. The first one disappeared without notice.
A large company is no more immune to that than a small one, or a community run site would be.
And at least with an individual or small company/group, you have someone to put your case to in that situation. Someone who can probably offer support and potentially lift the ban rather than the 'Get fucked' attitude of Facebook, Google, etc.
The server admin can kick you off just because they feel like it. At least with a large public platform there's the possibility of public exposure of bad behavior.
I don't direct people to my HN profile. I don't rely on that for identity. I don't try to build an audience on HN.
> If you use Facebook or Twitter, you're trusting Zuck, Elon and the folks working at Meta or Twitter.
These are ad-funded businesses, they have a financial incentive not to ban everyone they dislike. They also have a track record.
If I host a personal instance, it's good that I can ban anyone I want. If you host a public instance, you probably will try to enforce your rules somewhat consistently, else it's gonna be hard to earn the trust of your users.
I learned this from this talk:
A New HOPE (2022): ActivityPub Four Years Later: The Good, the Bad, and the Fedi
https://www.youtube.com/watch?v=vnciCz83t70
Which notably is a recent talk (July) but was given before the recent influx.
Because humans are moderating the small instances, they can do a better job than AI moderators, albeit with more work.
Moving from one Mastodon instance to another taking follows and followers is quite easy. Admins usually announce plans to close with enough time for people to make the move.
unless there is an election about who gets added to the list it s not even democratic. And the freedom of speech is generally protected, not subject to the whims of democracy
With mastodon/fediverse, you have a choice of numerous different 'twitters', rather than being stuck with the global twitter decision.
That's the part that interests me the most.
Which is why I (when I get around to it) will self-host my own ActivityPub instance (Mastodon? Pleroma? WriteFreely?).
That way, I decide what gets blocked and what doesn't. And since it's federated (assuming other sites will federate with me), I can still access the benefits of using fediverse resources, without being beholden to someone else to make such decisions for me.
From a longer-term perspective, ISTM that an ActiviyPub (AP) "user agent server" (UAS) which allowed individual users to federate with the resources of their choosing without having to "join" an extant instance, makes a lot of sense.
Rather the UAS would utilize your local (i.e., generated and managed/federated from your AP UAS instance) fediverse credentials and act as a proxy for you, communicating with the instances containing the resources you're interacting with and acting as storage/cache/server for the user's client apps.
This would make the fediverse even more decentralized and under user control. However, it would also make discoverability even harder.
Perhaps that's ActivityPub's killer app? A discovery app. Perhaps a protocol change to require a cryptographically signed summary/intro for each instance, along with with a searchable index?
It works a bit like "cell-phone number portability" in the US. You set your original account to point at your new account, and you tell your new account about your old account. All of your followers migrate to your new account automatically.
Of course, you could still get burnt if your host just disappears one day, because it can't point to your new account. So there's an incentive to use a well-run server with a good track record and a reasonable funding model, I guess.
Sorry for so many questions - do you know where can I read about the mechanics behind it?
This is absurd. Mastodon is a network of severs run by volunteers. Admins have every right choose what servers to federate with.
Besides, if you don't like "the whims of democracy", try freedom of speech under a dictatorship.
By giving you the choice of which server to use, you choose your moderation. If other people choose (or choose to delegate the choice) to block you, that's still democracy.
There are lists of servers with bad reputations that get passed around and form the beginnings of many servers’ blocklists, but there’s no central entity maintaining a single list that says which servers are or aren’t part of the fediverse.
It's actually a nice solution.
Freedom sometimes means no one wants to talk to you.
That kind of list needs pretty good moderation to avoid that. Also, it's not really that effective considering offenders can easily move servers.
Hell, it would be all to easy for adversarial group to just server hop and get a bunch of sites blocked purely because the mods of those sites are not on 24/7... or just allow for free speech
Am new admin for a friend-only instance; currently blocking seems to be done via user reports + checking in on the #Fediblock hashtag. It's all manual, and repeated admin work needs to be done per-instance, which is why the article says it's inefficient but a good trade-off (so you can choose the server you align with best).
Nah. De-federating is manual. Such an adversarial group would need to join up, start spamming, and keep on spamming long enough for their instance to be added to a blacklist, and for other admins to pull that blacklist in, without the admin ever acting in the meantime.
It's not like the fediverse is that big, either - people talk to each other.
Mastodon servers do often have a "federated timeline" that shows you all the posts from servers your instance knows about. People can also send you unsolicited messages by replying to you or @-mentioning you.
> Who is the person who has the final say which server may be part of the fediverse?
I don't think there's really a single person with the authority to ban someone from the whole fediverse; your instance admins always have the final say about whether their instance talks to another instance. But things like shared blocklists mean that, in the vast majority of unambiguous spam or illegal content cases, only one person actually has to respond to them directly.
You can deal with that by blocking them personally. No need for a guardian angel.
If your server's admin(?) makes a fediblock toot, that's more an statement that your own server has blocked someone and asks other servers to do the same. Admins of the other servers can always decide how to deal with that signal - ignore, issue the same block, boost the signal, etc. (Though there is probably some "peer-pressure"/reputation management at work, which makes their decisions not completely free)
So the "network-wide ban" really isn't. It's more a ban from the vast majority of the network - but there is no technical way to issue a genuine ban from all Mastodon instances, the way you can e.g. get banned from the entirety of Reddit.
That at least seems more democratic to me. Of course the usual pitfalls of democratic systems still apply: How to deal with bad-faith #fediblock signals? How to avoid hidden centralisation where certain instances gain more authority to issue blocks than others? etc...
If someone more experienced has some articles with more details about how the system works, that would be really cool.
Lower tier trackers will often adapt the cabal's decisions for various reasons. You end up with a situation where one staff member making a decision can end your membership on a dozen sites.
In the torrent world, it's pretty open. I expect the same to be the case on the fediverse, even if it's not publicly communicated. I'm sure server admins do communicate with each other in non-public forms.
The new Internet was like a telephone and could connect anybody to anybody.
This 'peer-to-peer' idea was good thing. It felt exciting and democratic. It remains a fantastic leap for humanity, worth defending to the death.
However, for some - a possibility once becomes a necessity unto the limit - that is to say the idea that;
"occasionally anyone could talk to anyone"
silently transformed into the idea that
"everyone should talk to everyone all the time"
whereas in fact
"most people don't want to talk to most people most of the time".
That is a problem at the intersection of people, technology and culture. As the earliest culture-shock pundits predicted, we still haven't really worked out what we want to use this technology for.
It would be best not to optimise too early, to leave our options open and not let any one model dominate development - which is the theme of resilience through diversity in the OP.
That's why I think if you don't self host the first server choice ("choose whatever you want it doens't matter") is extremely downplayed. Instances can block each other and if you end up on a "wrong one" (the peer pressure to avoid guilt by association is really really strong if you look around) then you have to move around, your name changes etc.
I do wonder how mastodon would handle influx of spammers. There is zero barrier to entry, and banning servers is ineffective if it is just an account doing the spamming.
A server that doesn't take responsibility for its users will soon find itself shouting into the void... as long as enough admins react.
Gmail is the largest source of spam because other mail systems won't ban it because it's too big.
How that helps ? It just makes impossible for small server admin to take vacation...
When comfortable, change to friends-of-friends.
And if you're going to grow past being comfortable with everyone on your server, have a backup admin you trust.
Google used to host the biggest xmpp server (google talk), and then they defederated from the rest of the network.
XMPP is but one of many protocols that were adopted by Big Tech, then once market share was achieved they took their users and walked away from the open internet
If Google ever speaks AP I won't be surprised to see it defederated from many instances. It will create a schism, and that's perfectly ok because the expectations are different.
You need a secondary link to your friends (email, phone, other server) so you can notify them when one of your vertices is taken down or another is added.
Yet of course that doesn't scale well to bigger social discussions, it's way too easy for malicious agents to destroy any chance for sensible wider discussion when there is no moderation present. HN/Reddit-esque "just let people downvote and push the downvoted stuff down" works somewhat in most cases, up until some group decides to vote brigade the conversation, and in wider discussion there is also the problem of "I downvote not because the comment is bad but because it disagrees with me.
There will be problems but there is evidence that it can work.
Mastodon doesn't have this kind of exception, so it suffers from a lot of conflict about who exactly the Bad People are and whether accepting messages from them causes you to become a Bad Person yourself.
Lacking a universal definition of bad actors does not prevent coordinated action to stop the bad actors.
I can disable my spam filter, and no one will know or care except for me (if I get a lot of spam). On the other hand, it is generally considered immoral to allow people to disable anti-Bad Person filters, because then two Bad People could use the network to communicate. That creates a lot of potential for drama, large public arguments, and network fracturing that isn't present in legacy systems.
People who use the service to plan murders? People who disagree with your politics? People who (accurately) quote a book written 200 years ago that uses a word you don't like? People whose skin color you don't like?
Even the "ban what is illegal by law", aside from some common stuff, varies from country to country. Anything above that is pretty much politics.
Like, it's fair to say "okay, this community is to discuss X, we don't do politics coz that's always divisive", but the moment mods try to play moral guide and decide what's wrong think and what isn't any bias they have will be magnified by the power they have.
You can pay someone to host a server for you, starting at $4/month. (Well, MastoHost is a little behind on signups this week, but they'll catch up.) If you do this, then you'll have total control over your moderation policy.
My instance is run by someone who mostly only blocks the absolute worst, including instances that are illegal under European law. I'm fine with those blocks! Other instances are focused on serving people who are regularly harassed, and those instances ban aggressively. There's even a "co-op" instance where users vote on policy.
I think it's worth remembering that for many users, a pleasant experience without harassment is a valuable service. I've moderated niche communities for years, and it has become obvious that if you let a handful of jerks abuse everyone else, then dozens of amazing community members will quit.
The solution to fragmented servers is not another self hosted disconnected server?
Then choosing to block other servers would be equivalent to blocking other people, which we agree is a good thing.
Now someone comes up with an adblock list: subscribe to it, and these servers will not bother you. It's probably not perfect, but it improves your experience a lot. If you don't like it, switch to another, or don't subscribe to any.
And now we are back where we started: you can choose to either subscribe to a policy (join an existing server) or be independent.
I've seen a couple of servers which only show the block list to users with accounts. These tend to be restricted-membership servers, ones which don't publish their member directory or local feed, either.
My impression was that many apply transitive banning? So that if I peer with a “bad” server A then good server B will block me even though my personal posts may be fine.
But it is, because the accounts are portable, so they can just move to another instance... it's an actually free market. In the end, you can always self-host if you literally can't agree with a single admin out there.
In the fediverse, after all, users who don't like that decision can always move to another instance, or set up their own - you can't do that on twitter, HN, Reddit, and the barrier to entry for setting up one's e-mail server is high enough that the capability to do that is only theoretical.
I don't think that's a desirable goal. It's up to individual services to host centralized forums and deal with the ensuing moderation issues.
The same people I follow on Twitter I follow on Mastodon and I feel they are being more honest on Mastodon. This might not be true.
But personally I haven't found the waves of Twitter users running from Musk to be part of that. For the most part they seem to be running into the fediverse and expecting the same culture that Twitter had.
As these sorts of waves tend to always go, I expect it to either fizzle out or to ruin the culture that everyone prior to them got to enjoy. Although maybe the fediverse will be a bit more resilient to this due to being less centralized.
It's a net-win for fediverse. But as I've said elsewhere, will not replace Twitter.
But the world also doesn't need a Twitter.
That kind of single "town square" doesn't work. Can't work. And just attracts blowhards and narcissists. And the world will be better without it.
A lot can hide in the unspoken assumptions behind the unqualified use of the idea of efficiency.
In this case the comparison of "efficiency" of Twitter vs. Mastodon is pretty meaningless. There are as many goals as there are users and moderators and (in Twitter's case) management and investors, eh? Of course, there are broad groupings and shared interests, but by and large the folks using Twitter and/or Mastodon have different goals.
It reminds me of comparisons between industrial farming and small-scale ecological farming: industrial farming is more efficient at creating massive amounts of some kinds of products, but it literally cannot produce any of some other kinds of products. (E.g. you can't get Alpine strawberries in the supermarket because they start to lose structural integrity the very moment you pick them. The only way to get them is to grow them close to your kitchen.)
- - - -
As a bit of a tangent, I like to reflect how evolution has no goals therefore speaking of the efficiency of living systems is meaningless. Evolution is a chemical tautology.
Quoting someone sad
Warum retweeten alle aus Twitter? Das ist, als ob man sich von einer toxischen Freundin trennt und dann nur noch von ihr erzählt
I hope they can handle it well.
It's nice that they allow off-migration and don't lock your community in.
https://masto.host/the-25-percent-commitment/
We will likely migrate to our own cloud setup once size gets stable.
However, you can have large general purpose instances which do not have to depend on federation to deliver the value of a network like twitter. Additionally, mastodon itself has grown by leaps and bounds in making the UX easier for the average user familiar to twitter/facebook.
I think its only a question of time before you have large general purpose mastodon instances where users dont have to grapple with federation and the UX becomes familiar for the layman. But at the same time there is space for smaller community focussed instances if someone wants to take their tribe with them.
I’m in Australia and a lot of people have moved from Twitter to a server called aus.social, so if I was going to tell friends or relatives to get on Mastodon I wouldn’t probably direct them to sign up to that specific Mastodon server, not just say “you should use Mastodon” and then expect them to work it out. That one has a lot more rules than the infosec ones though.
Locally made honey (i.e. not from South America) in Switzerland usually costs significantly more than 10€ :)
So I'm happy to hear Swiss prices are better. Though, having lived and worked there a summer, I guess it's still relatively cheap. The price of a coffee....
I don’t get to goto work and engineer essentially the same product 2007 engineered.
It’s a rehash of a rehash of an idea that will become another social landfill.
Really looking forward to the technology equivalent of another Star Wars or Batman trilogy!
I can imagine it being both more efficient and just better in various measures as well.
For starters just the ads and trackers must have a cost, which is paid on every visit.
As we've seen many times on the internet, cacheable content can be far more efficient than constantly updated content that needs refetched just to show you an ad impression. Google's web page speed tools will warn you about Google Ads non-cachable JS tracker.
Combine that with the incentive to show growth in 'user' metrics, which encourages both leniency on bots and intentionally addictive designs for non-bots and I can see an obvious case for twitter being less efficient.
Spritely is an effort to fix the problems that remain with ActivityPub.
If that's true, then your single sentence does more to communicate what Spritely wants to do than that entire website. Even searching for occurrences of "ActivityPub" on the site with Google¹ returns mostly 404s. It sure does not appear to be a living concern.
¹ https://www.google.com/search?q=site%3Aspritely.institute+%2...
Everyone who knows that a technology where decentralization is one of the requirements will know this, and they will also know that this is a required tradeoff when it comes to decentralization - no free lunch.
A lot of people skewer blockchain-based tech for ineffeciency, but not a lot of those people realize it's supposed to be like that.
I'm glad that people are at least starting to realize that a lot of decentralized tech has the prerequisite of forfeiting efficiency.
But my initial idea was to compare it two political issues recently in my country. That story was far too unrecognizable though.
The issues are, where a decentralized police, that was governed by local mayors, was replaced by a central police force, governed by national government. Centralization that was sold as being much cheaper and more efficient. The cheaper was a lie, but only because the conversion itself turned out far more difficult. The more efficient is true. But only for a certain part of efficient.
Previously, e.g. mayors could micro target issues that were directly plagueing a small neighborhood. Now, this is still possible, but has to move through layers of bureaucracy first. In many parts, i'm told, it's less efficient. Overall, the sum, however obviously operates more efficient. Instead of each of the thousands of municipalities having some HR, there now is only one. Same for IT, etc etc.
The other issue is exactly reversed: mental service for Youth (youth care) a rather large institute, was decentralized: moved to municipalities. The promise was that this would allow far better detailed care. It is a fiasco, because the national government never considered that it would be far more expensive. They even promised it would cut costs.
Why voters fall for this, and think that both promise can be true (decentralized being more expensive and saving money) is beyond me. But the effect of decentralization on effincy is clearly visible here too.
Write a twitter clone for a moderately sized community, lets say 10k with probably, 1-5% of community active at any given time. hell, even throw in activitypub if you want.
vs.
Write twitter.
I know which I'd rather do! Only one of the two is easier than rocket science! (haha)
Its inefficient because it hasn't needed to be efficient, but that will change in the future or someone else'll take over. Cascade issues are fixable. the end result is proven achievable (see telcos and the SMS systems that cover the world, global scale federated messaging exists and works!)
What you want to avoid is a celebrity on instance A posting and then users all over the fediverse pinging instance A to get that post. Instead, instance A sends the post to the fediverse and then for any given user to view the post, they ping their own instance. That allows server requirements to be more closely tied with the number of accounts on it and how heavily they use it rather than the popularity of any given account. Additionally, it makes it so servers only have to interact with their own users and other federated servers.
Of course the downside to this is that as the fediverse as a whole grows, the volume of the federated firehose increases, which still increases the minimum requirements (processing power, bandwidth, and storage alike).
There may be middle grounds here. For example, the instance I'm on (I have no idea how common this is) stores text posts from other servers but images from are not saved, it's the originating server that serves those. Additionally, there may be the possibility of deleting old posts from the fediverse that don't have any local replies, or something like that. There are issues with the latter, especially with old friends-only posts becoming impossible to access by looking at the public profile on the original server, but I could see that being solved by something like generating a readonly token for the other server or something like that.
This sentence at the very start already makes no sense. The above snippet makes as much sense as calling Web a network of nginx and Apache servers since nginx and Apache combined serve the majority of all web traffic.
There are services like Pleroma, Friendica, Misskey, Pixelfed, Write.as, PeerTube, and other Activity-Pub-speaking software which are NOT Mastodon or its forks. The author mentions these other pieces of software later, but pulling them them as "some part of mastodon" is at best a misnomer.
And it's been written about countless times already. See https://bofh.social/notice/APTBaDyBOOvNAzaeqO for a relatively fresh list.
At least I think we can hope that people will start to discover all the other things that the Fediverse contains ince they are there and start to use it.
Also, not sure if I'm correct, but from what I've read, the software seems to be quite influential in shaping the "de-facto standard" client/server API, at least for the "microblogging" part of the Fediverse?
Like, in theory, ActivityPub both specifies server/server and client/server interactions - however, the client/server part appears to be "underspecified and barely used" [1, 2]. e.g., AP doesn't say anything about how clients are authenticated or even how they can get notified of incoming messages/activities short of polling. This stuff is all specified in the Mastodon API.
So in practice, if I wanted to write a new client app for microblogging on the fediverse, my app would use the Mastodon API, not ActivityPub for reference. Is that correct?
Of course in the interest of avoiding centralisation, I'd agree that that's not a good state of affairs. It would be better if there was a real, widely-used standard for client/server and not just what the Mastodon devs think is useful. Though at least they cannot simply make changes to the API without convincing the majority of server admins to adopt those changes.
[1] https://socialhub.activitypub.rocks/t/activitypub-client-to-...
Can you elaborate on this? I really don't understand why you say this. In a GNU/Linux distro Linux provides the kernel, and GNU the userland tools. How can GNU/Linux be a subset of Linux? Your next sentence makes it clear you don't mean superset, so I'm lost at understanding what you mean.
Android is an example of one that does not. Although when someone describes an OS as Linux, to me it implies GNU/Linux, so I wouldn't really call Android Linux, even if it uses the kernel.
So Debian Linux is GNU/Linux. Proprietary app running on top of linux kernel in some embedded device is "just" Linux without GNU userspace.
It
It matters exactly nothing for the point of the post. It's entirely irrelevant for the story.
I know that "technically correct is the best kind of correct". And you are technically entirely correct here. But it still doesn't matter.
What is the point of making irrelevant comparisons is what I'm asking.
But energy and Co2 make the article even more convoluted. Bitcoin (PoW) even more so. And there's a major difference. In which PoW deliberately burns energy. Whereas the fediverse at least tries to burn as little as possible while remaining decentralized.
Unless you can incur a significant cost to these activities, e.g. requiring social capital (burning social capital with e.g. hate/disinformation), or physical (e.g. access requires boxes which cost some small fee and can be blacklisted)...
Perhaps some of mastodon inefficient behavior can be leveraged here?
It's easy to say that rules are 'censorship' in a vacuum, or when you're just talking with normal people, but when you get people intent on spamming or trolling or harassing, you quickly realize there's a reason human societies basically always end up with rules.
There are different views on nudity around the world. Nipples are normal where I'm from. There are different views and limitations on speech around the world. There are different views on copyright around the world. What's the term limit? There are different views on what constitutes animal abuse around the world. Try posting about the local bullfights on facebook. Or try posting videos from the cockfight from the cockfighting arena.
Different mastodon servers, and heck, maybe different fediverses, will fuel diversity in this regard, without One Set of Rules to Rule them All.
That said, in my opinion there are plenty of interesting people in the fediverse already. Personally I don't really care if the fediverse is widely adopted by celebrities and public figures or not.
Famously, Facebook's iOS app is over 100 MB[0] and for Android, their engineers hacked the Dalvik VM because they had so many class definitions they were overflowing a 5MB buffer just for the definitions - and then, staggeringly, they bragged about this obscenity.[1]
I can only imagine their backend systems are a similar nightmare, that leads to nonsense like FB building its own data centers to run their photos-and-ads crud app.
[0] https://www.cio.com/article/230043/why-is-facebooks-ios-app-...
[1] https://engineering.fb.com/2013/03/04/android/under-the-hood...
>I can only imagine their backend systems are a similar nightmare, that leads to nonsense like FB building its own data centers to run their photos-and-ads crud app.
Uh, of course they are running their own datacenters, they are one of biggest sites on net, outsourcing infrastructure to cloud makes no goddamn sense to them
Overstaffing and ship-the-org-chart syndrome leads to ginormous app bundles on the client and 7000 microservices on the server.
> outsourcing infrastructure to cloud
Paying for AWS or building a data center are not the only two options. There are many co-location providers, there are dedicated firms with the expertise to run a data center for you...
> Even if Mastodon were to be rewritten in Rust, tuned, and changed into a backend that can host thousands of users on a single Raspberry-Pi running on solar power, it still is inefficient. For one, because that backend would be even more efficient when employed in a centralized setup. And secondly because there is a lot of network overhead.
Reading these two paragraphs, I was struck by how similar it is to the argument against proof-of-work cryptocurrencies, only with the oppozite conclusions. Look how wasteful, they will say, look how inefficient, and slow, and non-green it is. It's rare to hear someone argue that yes, a technology is wasteful, and it is a good thing.