7 ms·

How to Kill a Decentralised Network

141 points by pimeys 3y ago

kaycey2022 3y ago
This was excellent, and I feel in my bones it's true. I couldn't really have articulated this better than you.
znpy 3y ago
This is the best piece of writing I’ve seen on hn in months. Really worth reading!
- nmcela 3y ago
  I agree. Especially the point about staying as a honest tool and not joining the growth-monopoly-privacy-theft -game at all resonated heavily with me. Also kudos to Kev Quirk. Meta in its entirety is incredibly immoral in every way.
- newlog 3y ago
  I came here just to share my appreciation to the writing. You beat me. 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.
  - rvz 3y ago
    Yes. There are those who have read up on the history of what 'Embrace, Extend, Extinguish' means and that the methods are different over the years and cleverly hidden by companies like Google, Meta and especially Microsoft but the strategy has always been the same. 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.
- spacebacon 3y ago
  here here
olivierduval 3y ago
It looks like we're back at some Open Source Licensing-like schema : an implementer should only be able to "extend" an open protocol if its extensions are themselves open ! 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
- YetAnotherNick 3y ago
  Things like these are a liability for open source volunteers. I wouldn't want to volunteer in open source for no payment if I could be sued for changing anything. It needs to define open. Are third party captchas open? Is requiring facebook account open? Is closed AI based spam control open?
  - 2Gkashmiri 3y ago
    When was the last time an open source maintainer was sued for changing something?
  - olivierduval 3y ago
    Actually, it would be exactly the same as some kind of (viral) Apache Licence: you can extend the protocol ONLY if your extension is open source too 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...
    - pessimizer 3y ago
      > why wouldn't it work for open protocol 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.
    - omoikane 3y ago
proactivesvcs 3y ago
I'm not sure the same thing is on the horizon for the Fedi because it's in a different position. If facebook did join and suddenly we have a huge number of new people to talk to, many servers will block facebook on sight, and so will many individual users. So they'll never interact or be part of it from day one. 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.
- lapcat 3y ago
  > If facebook did join and suddenly we have a huge number of new people to talk to, many servers will block facebook on sight, and so will many individual users. So they'll never interact or be part of it from day one. 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."
  - proactivesvcs 3y ago
    That split is only a split in one sense. If me, on Fosstodon which blocks facebook, wants to read your posts on a server that doesn't block, we are not impeded.
jeroenhd 3y ago
I don't get this. People are saying Meta will do the same thing when Threads or whatever they call it comes online. 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.
- logicprog 3y ago
  Tbf, Mastodon has about 7 million users, not a few ten thousands: https://mastodon-analytics.com/ https://mastodon-analytics.com/
  - rvz 3y ago
    7 million 'created' accounts throughout its entire existence in 8 years is hardly interesting or early days. The point of the OP still stands that "Nobody but a few geeks use Mastodon, Mastodon could never even handle all Twitter users moving over." 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 https://news.ycombinator.com/item?id=36012434 [1] https://masknetwork.medium.com/mask-network-acquires-pawoo-net-one-of-the-largest-mastodon-instances-5273504ba92d https://masknetwork.medium.com/mask-network-acquires-pawoo-n...
pessimizer 3y ago
Any journalist reading HN right now should see that this blog entry gives you material enough for a good, current story, maybe even a feature, and that its author in uniquely suited to comment because they were part of XMPP and in the middle of the OOXML debacle, so would make a great interview.
willmadden 3y ago
Embrace, extend, extinguish. Decent article.
- mariusor 3y ago
  The article is not decent because the premise "niche protocol is still niche after big player dumped using it" it's blatantly obvious. It's just another trite "X is dead, netcraft confirms it" juxtaposition of maybes and what ifs.
  - pessimizer 3y ago
    > "niche protocol is still niche after big player dumped using it" 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.
    - mariusor 3y ago
      The article is about how XMPP is a foreshadowing for what's going to happen to the Fediverse after Meta will enter the space. 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.
      - newlog 3y ago
        You miss the possibility that when all the compatibility problems start to happen, a possible organic adoption (by other than hobbyists) would be hampered. 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.
        mariusor 3y ago
        Another bunch of what ifs and maybes.
RcouF1uZ4gsC 3y ago
> And because there were far more Google talk users than "true XMPP" users, there was little room for "not caring about Google talk users". 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.
- Zash 3y ago
  But most of all, users care about talking to their friends. You could have all the features and amazing UX but if their friends are using tin cans and wet wire, that's what they will turn to. Network effect trumps all. Features and UX are marketing, which at most can help you bootstrap to the point where network effect kicks in.
zoobab 3y ago
From the No<OOXML> days, the only thing i regret is to have not sued Microsoft in court for having corrupted the whole process.
api 3y ago
Many words written and little mention of the big problem with decentralization: without an economic model you can’t afford the incredible amount of effort required to make software usable for non-techies. 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…”
diggernet 3y ago
One word summary: defederation Luckily for the centralized systems, the decentralized networks have a culture of defederation that is undermining their own mission.

141 points by pimeys 3y ago

kaycey2022 3y ago

This was excellent, and I feel in my bones it's true. I couldn't really have articulated this better than you.

znpy 3y ago

This is the best piece of writing I’ve seen on hn in months. Really worth reading!

nmcela 3y ago
I agree. Especially the point about staying as a honest tool and not joining the growth-monopoly-privacy-theft -game at all resonated heavily with me. Also kudos to Kev Quirk. Meta in its entirety is incredibly immoral in every way.
newlog 3y ago
I came here just to share my appreciation to the writing. You beat me. 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.
- rvz 3y ago
  Yes. There are those who have read up on the history of what 'Embrace, Extend, Extinguish' means and that the methods are different over the years and cleverly hidden by companies like Google, Meta and especially Microsoft but the strategy has always been the same. 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.
spacebacon 3y ago
here here

olivierduval 3y ago

It looks like we're back at some Open Source Licensing-like schema : an implementer should only be able to "extend" an open protocol if its extensions are themselves open ! 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

YetAnotherNick 3y ago
Things like these are a liability for open source volunteers. I wouldn't want to volunteer in open source for no payment if I could be sued for changing anything. It needs to define open. Are third party captchas open? Is requiring facebook account open? Is closed AI based spam control open?
- 2Gkashmiri 3y ago
  When was the last time an open source maintainer was sued for changing something?
- olivierduval 3y ago
  Actually, it would be exactly the same as some kind of (viral) Apache Licence: you can extend the protocol ONLY if your extension is open source too 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...
  - pessimizer 3y ago
    > why wouldn't it work for open protocol 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.
  - omoikane 3y ago

proactivesvcs 3y ago

I'm not sure the same thing is on the horizon for the Fedi because it's in a different position. If facebook did join and suddenly we have a huge number of new people to talk to, many servers will block facebook on sight, and so will many individual users. So they'll never interact or be part of it from day one. 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.

lapcat 3y ago
> If facebook did join and suddenly we have a huge number of new people to talk to, many servers will block facebook on sight, and so will many individual users. So they'll never interact or be part of it from day one. 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."
- proactivesvcs 3y ago
  That split is only a split in one sense. If me, on Fosstodon which blocks facebook, wants to read your posts on a server that doesn't block, we are not impeded.

jeroenhd 3y ago

I don't get this. People are saying Meta will do the same thing when Threads or whatever they call it comes online. 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.

logicprog 3y ago
Tbf, Mastodon has about 7 million users, not a few ten thousands: https://mastodon-analytics.com/ https://mastodon-analytics.com/
- rvz 3y ago
  7 million 'created' accounts throughout its entire existence in 8 years is hardly interesting or early days. The point of the OP still stands that "Nobody but a few geeks use Mastodon, Mastodon could never even handle all Twitter users moving over." 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 https://news.ycombinator.com/item?id=36012434 [1] https://masknetwork.medium.com/mask-network-acquires-pawoo-net-one-of-the-largest-mastodon-instances-5273504ba92d https://masknetwork.medium.com/mask-network-acquires-pawoo-n...

pessimizer 3y ago

Any journalist reading HN right now should see that this blog entry gives you material enough for a good, current story, maybe even a feature, and that its author in uniquely suited to comment because they were part of XMPP and in the middle of the OOXML debacle, so would make a great interview.

willmadden 3y ago

Embrace, extend, extinguish. Decent article.

mariusor 3y ago
The article is not decent because the premise "niche protocol is still niche after big player dumped using it" it's blatantly obvious. It's just another trite "X is dead, netcraft confirms it" juxtaposition of maybes and what ifs.
- pessimizer 3y ago
  > "niche protocol is still niche after big player dumped using it" 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.
  - mariusor 3y ago
    The article is about how XMPP is a foreshadowing for what's going to happen to the Fediverse after Meta will enter the space. 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.
    - newlog 3y ago
      You miss the possibility that when all the compatibility problems start to happen, a possible organic adoption (by other than hobbyists) would be hampered. 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.
      - mariusor 3y ago
        Another bunch of what ifs and maybes.

RcouF1uZ4gsC 3y ago

> And because there were far more Google talk users than "true XMPP" users, there was little room for "not caring about Google talk users". 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.

Zash 3y ago
But most of all, users care about talking to their friends. You could have all the features and amazing UX but if their friends are using tin cans and wet wire, that's what they will turn to. Network effect trumps all. Features and UX are marketing, which at most can help you bootstrap to the point where network effect kicks in.

zoobab 3y ago

From the No<OOXML> days, the only thing i regret is to have not sued Microsoft in court for having corrupted the whole process.

api 3y ago

Many words written and little mention of the big problem with decentralization: without an economic model you can’t afford the incredible amount of effort required to make software usable for non-techies. 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…”

diggernet 3y ago

One word summary: defederation Luckily for the centralized systems, the decentralized networks have a culture of defederation that is undermining their own mission.