Hey all, so I’ve been trying to embrace the fediverse life. My background - I’ve been on the internet since pre-WWW, so I’ve seen it all.
I think there’s a structural issue in the design of Lemmy, that’s still correctable now but won’t be if it gets much bigger. In short, I think we’re federating the wrong data.
For those of you who used USENET back in the early days, when your ISP maintained a local copy of it, I think you’ll pick up where I’m going with this fairly quickly. But I know there aren’t a ton of us graybeards so I’ll try to explain in detail.
As it’s currently implemented, the Fediverse allows for multiple identically named communities to exist. I believe this is a mistake. The fediverse should have one uniquely named community instance, and part of the atomic data exchanged through the federation should include the instance that “owns” the community and a list of moderators. Each member server of the Fediverse should maintain an identical list of communities, based on server federation. Just like USENET of yore.
This could also be the gateway into instance transference. If the instances are more in-sync, it will be easier to transfer either a user account or a community.
This would eliminate the largest pain point/learning curve that Lemmy has vs Reddit.
Open to thought. And I’ll admit this isn’t fully fleshed out, it was just something I was thinking about as I was driving home from work tonight
Lemmy is good, but it could be great.
My understanding is that if an instance suddenly dies, all the federated instances that subscribe to its communities will still have the text content because they store copies locally. So knowledge should not just go away. Media is a different story though.
I think new posts/comments in those communities would then not federate at all anymore since the host instance would not acknowledge them. So the communities turn into isolated local ones.
If the host instance comes back and the communities are re-created, they’ll be empty on the host instance but I think other instances won’t delete the old content unless explicitly requested.
How is this scalable and sustainable? If every instance caches every other instance then it’ll blow disk usage out of the water.