Yeah, ensuring availability over time requires dedicated infra. That's basically what it comes down to. Torrents for the most part lack dedicated providers ensuring file availability. Web seeds exist, but the uploader or the tracker needs to have the resources to back their torrents with bandwidth and storage. Other decentralized solutions, like say IPFS, don't solve the resources problem, because it's not technical, so although you can pay to have content "pinned" in place on IPFS, or you can pin it yourself, that "pinning" requires a server, running off electricity, using someone else's uplink to serve the content, all of which costs. If you don't have your own server, and don't pay someone else to pin it for you, it could easily fall off IPFS.

Syncthing could honestly help, I've thought about this a fair amount, although you'd still have the resources issues. Availability of content over syncthing or something like it would likely still be tied to popularity (how long are uploaders going to keep their syncthing folders full of specific content? how long will downloaders? In order for it to really work people would have to get in the habit of building out NAS's and putting their libraries on syncthing forever, basically). It still has some of the same basic issues with torrent, but the dynamicness is cool for sure.

[–] rdri@lemmy.world 2 points 9 months ago* (last edited 9 months ago)

Yes the availability will remain an issue but at least I imagine that solving other issues could make it less serious.

More specifically, the issue (a feature too but still) with torrents is how spread they are. It's difficult to know what is available and in what condition. There are dozens if not hundreds private trackers etc. This all makes it more likely for new torrents for the same content to be created multiple times, and overall seeding resources to be spread out across multiple versions of the same things. Some centralized public index might have helped everyone find things faster and prolong those things' availability as the result. What such an index might need to stay damage-proof and useful is unrelated to this discussion, but I imagine it might work as some blockchain and thus may not require much in terms of resources.

I didn't mean syncthing itself but some theoretical derivative that would have relevant features.

It would help to involve a kind of software infrastructure where users would choose how much resources (mostly disk space) they are willing to give in order to contribute to the overall availability of stuff.