we really need some kind of #mastodon feed mix controls. like, say I'm subscribed to someone who posts every 5 minutes, and I want to see their posts, but I don't want my entire feed to be just their posts. idk if that's a client level thing or a mastodon server level thing. i don't think it's #activitypub level
@2ck It's a client issue. Something like this could be done without changing Mastodon or ActivityPub. Currently, you might solve this using lists: put accounts which are way too verbose, but that you don't want to unfollow, in a special list and when you feel like reading from them, check that list, while maybe muting them from your home feed.
@2ck more complex home feed post selection processes could probably be implemented by clients, but something people in the fediverse really like is that posts are presented strictly in chronological order. So I don't think that there would be that much demand for something like that, as of now.
@junbird i think it could be server-side in part, though I'm thinking it's like the server providing hints about how some posts could be grouped beyond just being in the same thread. I took a stab at some basic clustering params:
https://qoto.org/@2ck/109126203727638798
@2ck No, all the server does is giving clients a list of posts (chronologically sorted) from each user. Each post also provides a list of replies. What an AP server does is actually really simple when it comes to providing content. However, I actually like your idea. That might be a pretty smart solution.
@junbird oh, yeah i know jack about what actual masto servers actually do. just think as a value-add, they could add annotations like clustering hints.
I want server-side so when i use tusky or the browser, i can have a similar experience in that regard. technically, it also seems a lighter lift if the client doesn't need to know the clustering rules and can just handle presentation.
i get the discomfort with opaque algos on the backend, but we're not out here pretending we know exactly what's running on servers we don't maintain, right? using annotations seems like a way to have the clustering without the server having the scope to quietly tailor your experience, since you can always turn off clustering client-side
