I have not once said that we need to get rid of servers, but I am saying that they could (should?) be used only as an proxy for the outbox/inbox. I’ve said already elsewhere, but it may make it easier to understand: the “ideal” model I have in mind is something like https://movim.eu, but with messages based around the ActivityStream vocabulary.
you really can’t get around this, even if you make every user handle their own stuff, every user will have their database and message queue.
Hard disagree, here. Tell me one system where I can take my domain and just swap the urls of the inbox/outbox. Mastodon lets you migrate your follower list and signals the redirect to your followers about your new actor ID, but you can not bring your data. But most importantly, the identity itself is not portable.
silo which holds all posts ever made, indexed to search? (…) that’s centralization
i don’t want my end device to connect to every instance under the sun.
Not every instance, but you’d be connecting to the outboxes from the people you follow. How is that different from, e.g, subscribing to a RSS feed?
my fedi instance db is around 30G, and im a single user instance which only sees posts from my follows
First: How the hell did you get this much data? :) I have an instance running for 4 years, with a bunch of relays, serving ~10 users and the DB has less than 4GB.
But to answer your question: If you are running a single-user instance, then you are already running a client, the only difference is that you are running on a remote machine which proxies everything for you. And how you deal with data wouldn’t change: just like you can delete old/stale data in Mastodon, you’d be able to delete or offload messages that are older than X days.
I have not once said that we need to get rid of servers, but I am saying that they could (should?) be used only as an proxy for the outbox/inbox. I’ve said already elsewhere, but it may make it easier to understand: the “ideal” model I have in mind is something like https://movim.eu, but with messages based around the ActivityStream vocabulary.
Why is it that a XMPP server can handle millions of concurrent users on a single box with 160GB RAM and 40 cores, yet Mastodon deployments for less than 10k active users have crazy expensive bills?
Hard disagree, here. Tell me one system where I can take my domain and just swap the urls of the inbox/outbox. Mastodon lets you migrate your follower list and signals the redirect to your followers about your new actor ID, but you can not bring your data. But most importantly, the identity itself is not portable.
You can have decentralized search indexes. Each server holds a bit of the index, but everyone gets to see the whole thing.
Not every instance, but you’d be connecting to the outboxes from the people you follow. How is that different from, e.g, subscribing to a RSS feed?
First: How the hell did you get this much data? :) I have an instance running for 4 years, with a bunch of relays, serving ~10 users and the DB has less than 4GB.
But to answer your question: If you are running a single-user instance, then you are already running a client, the only difference is that you are running on a remote machine which proxies everything for you. And how you deal with data wouldn’t change: just like you can delete old/stale data in Mastodon, you’d be able to delete or offload messages that are older than X days.