In order to avoid this restriction you would need a global instance discovery mechanism, which is extremely hard to implement without a central server that keeps a list of all instances in the network. And if you do implement instance discovery through a central server you really are losing the whole point of decentralization.
Additionally, it’s good that each instance does not federate with everyone else by default. If it did, it would have to process all activity and keep a local copy of all the content in the entire network. This would be insanely inefficient, and make it prohibitively expensive to run even a tiny instance with 1 user and no communities.
Decentralization isn’t useless if you can’t immediately see everything in the network, come on… We’re just spoiled by centralized services.
you would need a global instance discovery mechanism
I don’t think you do. Instances should merely reach out to other instances it’s federated with periodically to get a list of communities and some of their metadata. Ideally, they could ask all of those other instances to notify it when a community is added, modified, or deleted, and then store that metadata.
That should be pretty easy to implement, and maybe it already has, idk.
Sure, but this isn’t finding new instances, just new communities on known instances. Indeed, this is not difficult to implement. The reason it’s not done already is for resource economy. A lot of instances are already struggling to scale, making them process and store a lot more content with little value for most users of the instance isn’t feasible for a lot of servers right now.
A list of communities isn’t “a lot more content.” Just run it once daily and the problem is solved. Instances don’t need to store posts, just community names available on that instance.
If the issue is finding new instances, I think it’s fine for it to take some time the first time someone tries to find a community on that instance. But after that, it should immediately have a list of communities from that instance.
Ummmm… All you need is some bots on each instance that automatically will interact once with communities known on lemmyverse.net and boom, you have unlocked full federation for every user on those instances. You are not losing any privacy through that, it just skips the steps where user has to manually index a sublemmy before it federates and makes platform more usable.
Sure, but now this system has a dependency on the “centralized” lemmyverse.net service. And also your instance now has to receive and store a copy of almost the entire network’s content. Lots of instances are already struggling to sustain the load, this would make the problem even worse.
If a single instance decides that it can sustain the increased load and doesn’t mind depending on lemmyverse.net sure, nothing’s stopping them. But it shouldn’t be the default behavior for all instances.
In order to avoid this restriction you would need a global instance discovery mechanism, which is extremely hard to implement without a central server that keeps a list of all instances in the network. And if you do implement instance discovery through a central server you really are losing the whole point of decentralization.
Additionally, it’s good that each instance does not federate with everyone else by default. If it did, it would have to process all activity and keep a local copy of all the content in the entire network. This would be insanely inefficient, and make it prohibitively expensive to run even a tiny instance with 1 user and no communities.
Decentralization isn’t useless if you can’t immediately see everything in the network, come on… We’re just spoiled by centralized services.
I don’t think you do. Instances should merely reach out to other instances it’s federated with periodically to get a list of communities and some of their metadata. Ideally, they could ask all of those other instances to notify it when a community is added, modified, or deleted, and then store that metadata.
That should be pretty easy to implement, and maybe it already has, idk.
Sure, but this isn’t finding new instances, just new communities on known instances. Indeed, this is not difficult to implement. The reason it’s not done already is for resource economy. A lot of instances are already struggling to scale, making them process and store a lot more content with little value for most users of the instance isn’t feasible for a lot of servers right now.
A list of communities isn’t “a lot more content.” Just run it once daily and the problem is solved. Instances don’t need to store posts, just community names available on that instance.
If the issue is finding new instances, I think it’s fine for it to take some time the first time someone tries to find a community on that instance. But after that, it should immediately have a list of communities from that instance.
Ummmm… All you need is some bots on each instance that automatically will interact once with communities known on lemmyverse.net and boom, you have unlocked full federation for every user on those instances. You are not losing any privacy through that, it just skips the steps where user has to manually index a sublemmy before it federates and makes platform more usable.
Sure, but now this system has a dependency on the “centralized” lemmyverse.net service. And also your instance now has to receive and store a copy of almost the entire network’s content. Lots of instances are already struggling to sustain the load, this would make the problem even worse.
If a single instance decides that it can sustain the increased load and doesn’t mind depending on lemmyverse.net sure, nothing’s stopping them. But it shouldn’t be the default behavior for all instances.