Most of the problems in the current internet landscape is caused by the cost of centralized servers. What problems are stopping us from running the fediverse on a peer to peer torrent based network? I would assume latency, but couldn’t that be solved by larger pre caching in clients? Of course interaction and authentication should be handled centrally, but media sharing which is the largest strain on servers could be eased by clients sending media between each other. What am I missing? Torrenting seems to be such an elegant solution.
Centralized servers have a vested interest in serving content; they’re built around expecting freeloaders.
Torrent is designed around popularity. If a user seeks less popular content, it can easily be unavailable or slow. The fact that many expert seeders have leecher-hostile configurations exacerbates this.
To answer the original question: because it’s a worse experience for casual users, which drives them to a centralized model. There’s also the issue that by changing the networking technology, you cut off a huge percent of your target audience, creating an adoption hurdle; c.f. Gemini which, despite enormous user benefits, has foundered in creating a critical mass making it worthwhile for content creators to bother publishing on it. Mastodon and Lemmy have been successful in part because they don’t require users to download a bespoke app - they’re built on the web, which is centralized in design.
Change the Torrent content accessibility design issues, and you end up with something like Freenet, which ends up not only being slower (if, in the end, more reliable), but is far more resource hungry and unsuitable for mobile. IPFS might end up being a good alternative, but right now it’s still a bit technical for casual users, is slower than a centralized web, and can also be resource-heavy without client tuning.
Freenet has a good design and addressed some other Torrent weaknesses, such as lack of content anonymity, but also has some inherant flaws which may be unresolvable.