Source: https://lemmy.fediverse.observer/dailystats
Context: Reddit made a few controversial annoucements, feel free to have a look at !reddit@lemmy.world
For people wanting to discuss why some people focus on Lemmy’s growth, here is a recent thread from !asklemmy@lemmy.world :
We already have instances that go down or suffer from intermittent federation issues when lemmy.world gets a bit more active. The most conservative estimates are putting Reddit at 75 million DAU. If we get to 1% of that, you can bet that our current network would choke, badly.
Not only we need more instances, we also need to be a lot smarter about their organization and how to architect this network. I think we will only be able to grow larger if we make a more intentional separation between topic-based instances and “people-home” instances, so that we can have a better spread of the load.
The total federated bandwith is definatly a bottle neck we are starting to approach (ie what we see with .world and it overloading small instances). Not sure the solution here but I’m sure we can work past it without compromising on decentralisation.
0.19.6 is adding better parallelization
https://github.com/LemmyNet/lemmy/pull/4623
I am more and more convinced that we will need something like what I outlined here.
Good article. But the trend of the internet it to not use a browser but an app that often emulates a browser.
Which is still a client. I honestly don’t care if we are talking about a mobile app, a PWA, a browser extension, a SPA or a dedicated app: as long as the business logic goes to the edge and the server is a “mere” dumb pipe, we should be okay.
That would be better, but also put more cost on the users who have been spoiled by decades of someone else paying for incremental access to, storage of and processing of data.
Another article of mine arguing that modern devices are more than able to handle that.
I’ll read that but before I do, I want to point out that developers generally have the latest, most expensive tech and they generally build for that first.
We are not talking about complex tasks like video transcoding or ML training. ActivityPub is first and foremost, a messaging protocol. The most heavy thing that “must” live on the device is the user data. Sqlite can handle those workloads without a sweat. Old devices that can do XMPP group messaging should be more than able to do this.
I don’t know if it would help with load-balancing, but I feel hash tags would be better than communities.
This goes against the design of ActivityPub, which requires people to follow actors. A hashtag does not have a single name, so people would have to follow all servers and/or the servers would have to relay activities that are not originating from their actors. It is possible, clunky to implement.
Isn’t that just Mastodon and similar services? I prefer the community url scheme more that the hashtag scheme.
I agree that the hashtag scheme is bad. It attracts people that want to self-promote or bots.