Some people love saying Mastodon/OStatus is very scalable because of federation but let's be honest that's some bullshit. It's easy to split and multiply, but that's not scaling.
The birdsite is much easier to scale.
@CobaltVelvet that's not true, you're missing the point. OStatus scales better because it doesn't rely on a central large point of failure, the network can simply grow horizontally to accomodate more users
@sir ofc but scaling horizontally is still much harder in some cases
@CobaltVelvet how do you figure that applies to Mastodon?
@sir just look at my other toot there's a nice example.
The point was: being famous here is expensive. Being famous on the birdsite is a favor you make to them.
It's a problem and a feature of centralization
@sir Yes, that'd be a very good way to share the load.
@CobaltVelvet @sir that smells very bittorrent-like.
@sir @CobaltVelvet I haven't studied the protocol, but I thought the idea is that a toot only gets sent once to each federate, so broadcasting to millions of users is likely to boil down to broadcasting to (only) some thousands of federates. Having a single instance with a lot of users, on the other hand, requires horizontally scaling the service, but that is a "typical" web scaling problem that can be attacked with HA proxies, caching servers, etc?
@CobaltVelvet I guess that's true. But that also doesn't apply to most people, for whom famous only means at most ten thousand followers. Would be nice to devise a system (perhaps with PKI) that lets toots travel several hops to each follower instead of always coming from the tooter's instance.