Mastodon and Lemmy are turning into tiny DDoS botnets

stackdiary.com

15 points·skilled··10 comments

ActivityPub relies on a lot of POST requests being relayed from instance to instance in a full mesh style of networking. I have a lot of doubts about its scale, and also it's reliability in federating important actions like moderators removing offensive content. I'm going to need to read through the spec to confirm if this issue has been addressed, but there really needs to be a way to send multiple activities with one POST request, guidelines for retrying activities to ensure eventual consistency, and to accommodate message queues.

They have ways of bundling already, not all of them use it, also hes not talking about POST, he's talking about GET requests after a share is made, i'd imagine it's to cache the page or get info or something, when somethings shared. It's not the activitypub posting or something like that.

I'm glad to hear that bundling is already a thing. Combining that with a message queue will help a lot to ensure consistency.

> Again, I don't know what the exact purpose of this (probably to get the open graph data?) is

Yes, it's for the link preview. Each server that sees the link wants to generate a preview card for it. ActivityPub requests themselves are only sent between the servers that participate in the network.

Yes… let your hatred of OpenSocial and ActivityPub flow, you supporters of Big Social!

He has a legitimate point. Instead of doing it the exact same moment maybe they have a random 60s counter before fetching. Spread out the 4000+ requests over a minute. To addd another minute if the server numbers rise every 10k?

I mean hes also complaining when it has 50 requests...

What indication did the author give that they hate OpenSocial or ActivityPub?

Pointing out problems is hatred. Right.

https://jort.link/

Reply on news.ycombinator.com