My point was that there is no need to replicate everything everywhere. If the data is replicated a cross 5 instances per region for instance, it’s enough for replication needs. If you self host lemmy and subscribes to large communities on your instance, you can quickly overload your server. We need activity pub to be more lightweight if we want smaller instances to thrive.
They get hated on because :
they inspect packets. They terminate the TLS sessions at their servers and reencrypt to forward to the backend. This allows them to analyze the data to spot spam, optimize compression and such
they are used everywhere. If they go down, 30% of the internet goes with them.
My main gripe with ActivityPub is that the infrastructure basically replicates 1-to-1 across subscribed instances. It means that as lemmy grows, servers will require more and more storage to keep up. For now, it’s fine since we’re under a few TB of content on the platform.
If lemmy were to be as popular as reddit, we’d reach the dozens if not hundreds of TB of storage required. Not everyone has the money to build such a homelab or rent data center servers of that caliber.
ActivityPub in it’s current state is nothing but replicated centralization, not a full decentralized protocol. We’d probably need a different database system that handles cross region clustering and sharing to scale it up.
Playing devil’s advocate here but there could be legitimate reasons to prevent features of an app if you don’t give the permissions.
Things like professional type apps that need geolocation to work (geofencing, photo geoloc) or because x big shot client wants to track their employees and you’re just forced to accept that unless you want to declare bankruptcy.
Definitely is a very hostile pattern though and there’s no reason for meta to do this shit…