Comment

Comment on How to Kill a Decentralised Network (such as the Fediverse)

Oh, I read it when it came out back in June. Many times, as it kept being shared as an explanation of the first Threads backlash.

It's full of incorrect assessments and false equivalences.

Threads doesn't really have the volume (yet) to subsume ActivityPub. The process it describes for standards drifting towards the corporate actor doesn't apply to ActivityPub, whcih is engineered from the ground up to support multiple apps with differnent functionality (hence me writing this in Kbin and others reading it in Lemmy and being able to link it and follow it from Mastodon), the article only acknowledges that XMPP survived and kept on going at the very end as a throaway and doesn't justify how it "never recovered" and, like I said, it doesn't acknowledge the real reasons Talk and every Google successor to Talk struggled and collapsed.

So yes, I read it. Past the headline and everything. I just didn't take it at face value. This piece keeps getting shared because XMPP wasn't ever that big to begin with, so this sounds erudite and informed while the similar arguments being made at the time about SMTP and RSS were more obviously identifiable as being wrong for the same reasons.

source

Sort:hotnew top

sudneo@lemmy.world ⁨11⁩ ⁨months⁩ ago

which is engineered from the ground up to support multiple apps with differnent functionality (hence me writing this in Kbin and others reading it in Lemmy and being able to link it and follow it from Mastodon)

I mean that’s basically what every protocol is. ActivityPub abstracts concepts, that apps implement in their own way (for example the concept of group). If you manage to deliver changes, even improvements, to the protocol, apps need to keep up and comply with it. This is what means “drifting towards the corporate actor”. I propose changes to the protocol to a rate that only me (the corporate actor) can keep up with. This way only my users will have certain features and eventually some apps will become incompatible with the recent version(s) of the protocol.

source
- MudMan@kbin.social ⁨11⁩ ⁨months⁩ ago
  That is already how ActivityPub apps work.
  
  It's also not what happened to XMPP and, interestingly, not what the article claims happened to XMPP, even. You'll note in the postmortem about it the recollection is that Google was too slow to adopt features and fix bugs, not the other way around.
  
  I guess once you get enough confirmation bias in play you can embrace, expand and extinguish both by doing that and the opposite of that.
  
  source
  - 0x4E4F@sh.itjust.works ⁨11⁩ ⁨months⁩ ago
    
    You’ll note in the postmortem about it the recollection is that Google was too slow to adopt features and fix bugs, not the other way around.
    
    You still fall behind on compatibility with the original protocol. Doesn’t matter if you pull up or down, it still breaks compatibilty.
    
    source
    MudMan@kbin.social ⁨11⁩ ⁨months⁩ ago
    It absolutely matters under ActivityPub because, as I said earlier, it comes down to the client to manage the incoming packets. If Meta is out of date with the protocol but the rest of the federation is not (and retains backwards compatibility at all), then everybody else gets nicely formatted Meta content and they get garbled stuff.
    
    It's only relevant if we get garbled stuff and they get nicely formatted stuff. Which should be entirely avoidable if they "pull down".
    
    So no, not the same. And crucially people are still misquoting the article and the article is still misrepresenting the so-called "EEE" strategy.
    
    source
    -> View More Comments
    amki@feddit.de ⁨11⁩ ⁨months⁩ ago
    This is a core problem of distributed systems though. Signal even cites this as their reason to not federate with anyone.
    
    Once you get decentralization going you need everyone to stay kind of up to date or stuff will just not work.
    
    source
- amki@feddit.de ⁨11⁩ ⁨months⁩ ago
  It is not. Discord’s protocol has been tailormade to suit Discord and the developers will not give a single thought about keeping it stable because only the Discord server&client are meant to use it.
  
  source
  - sudneo@lemmy.world ⁨11⁩ ⁨months⁩ ago
    This misses the point in my opinion. The point of a protocol is to establish a set of rules that need to be followed, that’s it. After this, it can be managed in many ways, it can be open or it can be closed, etc. The fact that ActivityPub is “engineered from the ground up to support multiple apps with different functionality” it’s because ActivityPub is an open protocol. Every protocol is designed to support whoever implements it. This doesn’t have any inherent “the protocol (changes) will suit everyone” or “everyone will be able to keep up with it” property, though. If changes to a protocol happen very fast, apps that are compatible today - and can be compatible tomorrow too - still need to implemented those changes, or at some point they will not be compliant anymore. This is not because the protocol loses the property of supporting multiple apps, but because a protocol still needs to be implemented, which is responsibility of the consumers, which requires time.
    
    So my point was to challenge OC perspective that since ActivityPub is designed to support multiple apps, then there is no risk that it gets messed up and breaks compatibility with those apps (because it’s generic) due to - in this case -Threads influence. This is just nonsense, in my opinion.
    
    source