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?
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.
> 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.
10 comments
[ 3.6 ms ] story [ 35.6 ms ] threadYes, 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.