hhmx.de

Föderation EN Do 01.05.2025 15:54:11

I did some thinking about how boosting could be improved with smarter clients, a-la TCP congestion protocols

sethmlarson.dev/better-boostin

Föderation EN Do 01.05.2025 19:00:41

@joeneXtra @andypiper ooo thank you for sharing this!

Föderation EN Do 01.05.2025 19:05:48

@sethmlarson That's an interesting concept. One problem might be to determine the "time of last boost", because that's not a part of Mastodon's status entity, and thus might have to rely on incomplete information the client happens to have about the status.

Other than that, "delayed boosts" and a "boosting queue" that gets evenly spread in time are great ideas!

Föderation EN Do 01.05.2025 19:10:56

@bocops 100%, funnily enough I actually wanted to put a graph in this post showing how a typical post of mine behaves by graphing boosts over time relative to post time and found that indeed the "time of a boost" data is difficult to get...

Föderation EN Do 01.05.2025 19:17:11

@sethmlarson This should all be server side.

Clients don't know if/when they will run (think mobile apps with spotty network connectivity, or browser tabs that could be closed at any moment) so a client can't meaningfully "delay" a boost.

My server already knows what's in my timeline, so it could "drip-feed" boosts into my timeline over the day, or aggregate multiple boosts of the same post into a single boost-post in my timeline.

Föderation EN Do 01.05.2025 19:31:23

@nikclayton something like this would be basically necessary if we made boost behavior "better" (ie, boosted posts are more likely to meet new eyeballs).

Föderation EN Do 01.05.2025 19:30:35

@nikclayton Yeah that could work, all of the theory-crafting I was doing didn't involve servers somewhat deliberately because I didn't want to wait for servers to "choose" for me :)