federico3

joined 4 years ago
[–] federico3@lemmy.ml 12 points 2 years ago

If say lemmy.ml is down, what’s the point of other servers existing, if most of the content and users are here?

There is no replication and failover so the problem is not solved.

blockchain technology

Urgh, no way. Replication and some basic message signing would be enough.

[–] federico3@lemmy.ml 2 points 2 years ago

you could design the system in a way that one instance is mirrored and load-balanced across different hosters

That's exactly what I meant. Horizontal replication shares a lot of building blocks with federation. NNTP had peering/replication and worked quite well for a protocol designed in 1986.

[–] federico3@lemmy.ml 13 points 2 years ago (16 children)

Indeed. If a big instance like lemmy.ml was to be shut down all the communities would be lost. This is simply not sustainable. Why would users put effort building a community if it could be gone at any time?

[–] federico3@lemmy.ml 3 points 2 years ago (2 children)

What you are describing is just a form of "remote following", which is merely local caching of some content from another instance. As you wrote, each @gaming is an entirely independent community, even if the moderators are the same people across multiple servers. If an instance is shut down the community is gone. If the instance decides to throttle access and start charging money users have to pay or abandon the community. In short, this is not a significantly better user experience than traditional online forums. I'd rather have real federation.

[–] federico3@lemmy.ml 7 points 2 years ago (6 children)

Unfortunately that breaks the concept of federation. I expected servers with good relationships with each other to replicate posts, otherwise what's the point of federation?

[–] federico3@lemmy.ml 0 points 2 years ago (1 children)

The protocol could require "dual-homing" user accounts, where each account is automatically replicated on 2 different instances without need for hacks and workarounds. That would prevent users from losing their account if an instance is shut down, and also make it easy to migrate to a new instance without losing followers etc. The clients following your account always check for updates on both instances and if you move one of your accounts they update automatically.

(This would not create significant additional load on the network: your toots are already being replicated on all instances where you have followers.)

IMHO, you can only provide tools

No, tools are rarely "neutral". They encourage or discourage workflows and behaviors.

[–] federico3@lemmy.ml 0 points 2 years ago (3 children)

It's not that simple. ActivityPub is at risk of centralization, just like email. There are no built-in protections against centralization or EEE (Embrace, Extend, Extinguish). Furthermore, Mastodon makes it difficult to migrate accounts, especially from an instance that is unreachable or just disabled the export function.

Unfortunately locking users into a platform is extremely valuable because they can be shown ads, used for data mining, manipulation (like Cambridge Analytica). ActivityPub is not automatically immune to all of this.

The comparison with IRC is not very meaningful: moving from one server to another is much easier because IRC users don't lose followers, bookmarks, posts, etc.

[–] federico3@lemmy.ml 1 points 2 years ago* (last edited 2 years ago)

Flatpak, snap and docker are the problem.

[–] federico3@lemmy.ml 1 points 2 years ago* (last edited 2 years ago)

Not really. Signal is a walled garden but Matrix has a lot of a privacy issues.

[–] federico3@lemmy.ml 1 points 3 years ago

"Free" as in "unpaid labor" or "free to proprietize my work"

view more: ‹ prev next ›