If someone could create something where the servers only did a bit of work and the user base kept a set couple gigs each of the most recent clips they watch stored on their devices and it was all streamed/shared like a torrent would be to everyone else who was wanting to watch them, it would significantly cut down on costs.
Their wouldn’t be any long time storage of old clips for people to go back and find (although at more costs for server/bandwidth it could be, since those clips would be watched much less) but that’s not usually what people doomscroll through tik tok for.
But getting a server to just link up videos from thousands of other devices to arrange torrent uploads/downloads of clips could really reduce overhead.
I had to look that up, but I’d say yes, a lot like it, but there would probably have to be a bit more legwork from a central server for assigning out upcoming videos for each device. Something that keeps track of trends and stuff instead of just manual requests by the end user.
Yeah, some kind of coordination server, or servers? That sounds like a process that wouldn’t necessarily have to be super centralized, maybe it could be just a local thing, “show me what nodes near me are viewing”.
I think a problem with this approach is that doomscrolling is biased heavily towards leechers and people who have data quotas (e.g. phones using mobile data). Someone has to foot the bill of serving up all that content, and it probably doesn’t make sense that relatively unstable connections like cell phones are serving it. And there’s probably also latency costs to worry about too. People aren’t going to doomscroll as long if they have to wait 5 seconds between reels.
Most people have unlimited data on phones nowadays, and same for over wifi, so I don’t see a bit of upload time capping most people. As to the 5 seconds of buffer time, I was thinking the system would try to keep like the next 20 upcoming videos pre-loaded on your device in order to eliminate that. Not doing it on demand each time you’d swipe.
If someone could create something where the servers only did a bit of work and the user base kept a set couple gigs each of the most recent clips they watch stored on their devices and it was all streamed/shared like a torrent would be to everyone else who was wanting to watch them, it would significantly cut down on costs.
Their wouldn’t be any long time storage of old clips for people to go back and find (although at more costs for server/bandwidth it could be, since those clips would be watched much less) but that’s not usually what people doomscroll through tik tok for.
But getting a server to just link up videos from thousands of other devices to arrange torrent uploads/downloads of clips could really reduce overhead.
That’s basically ipfs I think
I had to look that up, but I’d say yes, a lot like it, but there would probably have to be a bit more legwork from a central server for assigning out upcoming videos for each device. Something that keeps track of trends and stuff instead of just manual requests by the end user.
Yeah, some kind of coordination server, or servers? That sounds like a process that wouldn’t necessarily have to be super centralized, maybe it could be just a local thing, “show me what nodes near me are viewing”.
I think a problem with this approach is that doomscrolling is biased heavily towards leechers and people who have data quotas (e.g. phones using mobile data). Someone has to foot the bill of serving up all that content, and it probably doesn’t make sense that relatively unstable connections like cell phones are serving it. And there’s probably also latency costs to worry about too. People aren’t going to doomscroll as long if they have to wait 5 seconds between reels.
Most people have unlimited data on phones nowadays, and same for over wifi, so I don’t see a bit of upload time capping most people. As to the 5 seconds of buffer time, I was thinking the system would try to keep like the next 20 upcoming videos pre-loaded on your device in order to eliminate that. Not doing it on demand each time you’d swipe.