How to Kill a Decentralised Network
ploum.net
ploum.net
Let's say that over the next few months facebook add some cool new features to ActivityPub which don't work very well on Mastodon, are completely broken on Pixelfed and Lemmy's dev is actively against. These features are things that facebook users are accustomed to, and some pre-facebook Fedizens really like the idea of.
If facebook then decide to cease federating, except with instagram, twitter and tumblr, me as a Mastodon and Pixelfed user simply doesn't care. They can vanish for all I care. I have a bountiful network of pre-facebook Fedi people already. I'm sure Shiny New Features will lure some people back to those abusive platforms, but a substantial majority of the current active user base on the Fedi are there because we despise being abused by social media companies. The developer communities behind the software that I use will stick with their parts of the spec, forking it if need be.
We'll do without the shiny features for now, and we have plenty of friends, or at the very least, plenty of friends yet to find on the platform. The Fedi already has its critical mass of users, developer communities, and servers, using and creating the platforms for reasons which are purposefully against the abusive social media's aims. They will carry the current spec off on their own if they need to.
Well, the debate about whether to preemptively block Meta, which has been raging among Mastodon users, is precisely the context in which the article was written: "if Meta joins the Fediverse, Meta will be the only one winning. In fact, reactions show that they are already winning: the Fediverse is split between blocking Meta or not."
There's a suggestion made by some, myself included, that instances which fail to block Facebook should also be blocked. That is, costs of defection are increased.
I'm not sure how it can be done legally binding... Maybe with IP on the protocol himself (as a technology) or with TradeMark (any reference to the protocol is forbidden without strong conformance to the protocol). OSS really need to level up the legal game
It already worked (more or less) for software... why wouldn't it work for open protocol ? And it would avoid the "embrasse, extend, extinct" strategy used by Microsoft or Google to transform an open protocol into a closed proprietary using extension and the size of their user base...
And if there were a license that says "you must also share all your data", it would imply compliant services can not hold private data, which makes it undesirable to companies and users alike.
Because it doesn't really make any sense. I can make my software compatible with yours completely against your will, and I can extend my software as I wish.
Really worth reading!
What a wonderful read. Very insightful. I had already read about this strategy before, but the way it is explained, the first-hand examples, and the depiction of what the true goals of decentralized networks should be is extremely clear and insightful.
It's very counter-intuitive to understand how adoption can go against the success of a software project, and it's also very easy to think that adoption is the goal, when in reality, maintaining (or organically evolving) the ethos of the software is the actual goal.
I wish I knew more people interested in reading what this writing talks about.
That applies here in this case. But unfortunately, you have Mastodon cheerleaders here who are asking for 'adoption' without knowing the trade-off of instance centralization and continue to blindly accept companies like Meta as a win for 'adoption'; unbeknownst to them signing up and bound to NDAs and getting themselves extinguished as Threads will add more features unavailable on many instances and then it becomes the biggest ActivityPub-compatible network that the users will use.
The warning from this article is totally accurate and Meta is already winning before many techies here have started to realize this.
Also kudos to Kev Quirk. Meta in its entirety is incredibly immoral in every way.
Computers are by default very hard to use. Techies don’t notice this because they know how to use computers well.
As a result techies think software is done when they can use it. It’s nowhere near done. Not even close.
Getting software working is often like 5% of the effort. The other 95% is polishing, and polishing, and going through 50 design iterations, and polishing more, and supporting a bunch of platforms with all their quirks, and polishing…
All companies have to do is create an easier to use alternative and a migration path. Then when enough people migrate, close the path. Done.
It’s all about UX.
I like the fediverse and use it, but if I didn’t understand how it worked I’d find it very confusing. Understanding how it works is easy when you have a network protocol background but without that the way users and instances and federation work is not obvious at all. That alone is “full stop” for the vast majority of people.
“I made an account… why can’t I log in?” “You aren’t on the right instance.” “Huh? What’s an instance?” “It’s a server that networks with other mastodon servers…” “So I need an instance?” “No you just need to go to the right one.” “How do I find it?” “It’s the one you made an account on.” “But I made my account on Mastodon…”
Luckily for the centralized systems, the decentralized networks have a culture of defederation that is undermining their own mission.
I'm pretty sure it's more: "growing niche protocol collapsed to near invisibility after big player reimplemented it, extended it, then broke compatibility"
If we're making up quotes, they might as well reflect the article.
What I'm saying is that hobbyists (like me) that develop fediverse software will probably be there _after_ meta gets bored and finds reasons to drop out, which is pretty much what happened to XMPP. And that it does not mean the death of anything.
At the same time, hobbyists that could be joining the network because of their initial ethos, would also refrain from joining given the blur and dilution of its values.
Sounds like it was already centralized even before Google stopped supporting XMPP.
The truth is that users do t really care about decentralization, they care about features, ease of use, and quality of implementation.
Once a majority of the uses move single implementation of the protocol, the single implementation protocol itself becomes more important than the standard.
Looking at the networks themselves, there are massive differences in user numbers. Nobody but a few geeks used XMPP. Nobody but a few geeks use Mastodon, Mastodon could never even handle all Twitter users moving over. Then there's a huge influx of normal people into the network because a big company experiments with being friendly with open standards. A system used by tend of thousands is joined by millions.
When those millions move away, the old protocol isn't killed. We're still in the situation that nobody but a bunch of geeks use XMPP. Some of those geeks don't care enough and stick with what works (though I doubt they've landed on Google Chat).
Those millions brought in by Google and Facebook didn't care about the federated nature of an extensible messaging protocol, they wanted to talk to their friends. Some of their friends had weird usernames that didn't end in gmail.com or facebook.com, that's the normal user experience.
XMPP sucked as a messaging tool and even today it still sucks because there are only a few clients out there that actually support basic group chat that everyone expects these days. Mastodon sucks because of usability problems (try following a user from a linked thread, the weird popup system is user hostile at the very least).
When some big company joins your network, you'll have to find a way to keep up with it or it'll leave again. Threads will join Mastodon, probably stir up some drama, and then defederate. Mastodon will then quickly go back to its natural state, a slightly smaller network of servers like it is now.
If your users jump ship because the competition made a better version, that's not killing your protocol. That's highlighting the problems with your current protocol, and your dependence on someone else's customer base to keep your project popular.
Even when Meta connects Threads to ActivityPub, it is a magnificent win for centralization and a complete nightmare for Mastodon and 'federation' and there is nothing they can do to stop it other than just de-federate and stay irrelevant.
This post exactly highlights the same thing on what happened to XMPP, but worse. Especially even when admins of other large Mastodon instances can be corrupted by larger companies like Meta with money, NDAs and contracts to allow these instances to federate with Meta. This is the typical 'Embrace, Extend, Extinguish' strategy all over again.
As I said before: "Instagram's success on ActivityPub is also Mastodon's nightmare and it tends to more centralization." [0] A paradox that is unavoidable for Mastodon without accounting for its UX failures and risking more centralization to the point where Meta is already winning after when other companies can still 'buy' out admins of very large instances. [1]
[0] https://news.ycombinator.com/item?id=36012434
[1] https://masknetwork.medium.com/mask-network-acquires-pawoo-n...
Interesting is a matter of perspective, but I've noticed over weeks and months that your HN account "rvz" has been constantly commenting about Mastodon, so you appear to be very interested, despite your claim to the contrary. ;-)
> early days
logicprog didn't say it was early days.
> Nobody but a few geeks use Mastodon
The issue is with "few". Millions of people aren't a few. Perhaps Mastodon users are all or mostly "geeks", i.e., technically sophisticated, but let's not pretend there aren't millions of users. Yes, the number is small in comparison to Twitter and Instagram, but it's large in absolute terms.
> Mastodon could never even handle all Twitter users moving over.
This is true.
Some Twitter migrants to Mastodon want to bring everyone along. Others, such as myself, are happy to leave all of the celebrities, politicians, and trolls behind. My community, Apple developers, for whom I originally joined Twitter back in the day, have almost all migrated from Twitter to Mastodon, and my Mastodon feed is as active as my Twitter feed ever was, while my Twitter feed is now largely dead.
Is there evidence that there are over a million regular users? I'd actually bet there aren't. I'd bet it wasn't even close until the twitter media panic.
However, there may be 7 million users, but about 2 million still use the platform at all. Two million people is a lot of people, but compared to Twitter's 240 million user accounts and Instagram's 2350 million user accounts the Fediverse is a spec of dust.
Yes, and instead of using "true XMPP" to talk to their friends and contribute to its growth, Google diverted that flow of new users into its own product. Sure the growth would have been far slower without Google, but this way Google managed to steal a few users that would have joined XMPP even without Google, on the false promise of compatibility.
Instead of slow, healthy growth, what happened was fast, parasitic growth, and on top of it all it added a maintenance burden for XMPP developers trying to maintain compatibility with Google.
You say XMPP was small - that's the point! To kill competitors before they become a threat.
WhatsApp picked XMPP as a basis for their messaging infrastructure when it started out. It never even bothered to federate with XMPP, but the technology still worked. XMPP's miserable state isn't some evil Google plot, it's a lack of user interest and a failure to effectively market the protocol. Unlike previous attempts, WhatsApp actually took off, easily overtaking Google's own fledgling messaging service, because it knew what people wanted: "chat without SMS charges". This left some space in other countries where SMS was essentially free, but nobody managed to capture interest in XMPP there either.
I know one XMPP client that everyone agrees is somewhat usable, Conversations, and it looks like development and design stopped somewhere around 2014. Absolutely nothing stood in the way of XMPP taking the place that Line, Telegram, Signal, WhatsApp, and all the others, but there was simply no interest in making a good product out of a federated service.
Matrix is trying again, with big companies using it instead of IRC, and it's struggling to survive in a world where Discord looks better and works better.