Attached: 1 image
hello, world!
The Rust Foundation is excited to join you on Mastodon. Going forward, we plan to exclusively share social media content here, on Bluesky, and on LinkedIn. 🦀
RustConf is also on Mastodon at @rustconf@hachyderm.io. Follow them if you're interested in the many updates about RustConf coming up!
Client side is completely out of the question here. You can’t just send millions of posts to every client hoping that they will order it themselves. Bluesky has a well thought out architecture that does that through server side feed gens, that can be decoupled from the App itself.
I know that, in my mind the algorythm or maybe better to said the algorythm’s parameters run on client side, no need to create a Fediverse instance on client side.
We can already to something a bit like that with word filters.
I’m not sure you understand how a suggestion algorithm works, it’s not something that can be run locally without having to then load and process thousands of messages. There is no aspect of that that could be run locally while simultaneously not having to load all the messages you’re attempting to sort and make recommendations on.
It is not at all equivalent to a word filter, a word filter is not attempting to find something, it’s a filter that applies to things that are already being loaded regardless. When a post comes in the word filter looks to see whether or not it has a word that needs to be filtered. An algorithm needs to find new content out of all the content available and then attempt to determine whether or not it’s something that should recommend meaning the client would have to load thousands upon thousands of random posts from the Federalists in order to make that algorithm work
Client side is completely out of the question here. You can’t just send millions of posts to every client hoping that they will order it themselves. Bluesky has a well thought out architecture that does that through server side feed gens, that can be decoupled from the App itself.
https://docs.bsky.app/docs/advanced-guides/federation-architecture
I know that, in my mind the algorythm or maybe better to said the algorythm’s parameters run on client side, no need to create a Fediverse instance on client side. We can already to something a bit like that with word filters.
I’m not sure you understand how a suggestion algorithm works, it’s not something that can be run locally without having to then load and process thousands of messages. There is no aspect of that that could be run locally while simultaneously not having to load all the messages you’re attempting to sort and make recommendations on.
It is not at all equivalent to a word filter, a word filter is not attempting to find something, it’s a filter that applies to things that are already being loaded regardless. When a post comes in the word filter looks to see whether or not it has a word that needs to be filtered. An algorithm needs to find new content out of all the content available and then attempt to determine whether or not it’s something that should recommend meaning the client would have to load thousands upon thousands of random posts from the Federalists in order to make that algorithm work