Peerweb: Decentralized website hosting via WebTorrent

Original Article ↗ Hacker News Discussion ↗

Centralization vs “decentralized” hosting

  • Several commenters question why files are uploaded to peerweb.lol and shared via that domain, arguing this is still a single point of failure (what if the site or tracker goes down?).
  • They note that p2p storage/distribution is “solved” (torrents, IPFS), but censorship‑resistant addressing and discovery without centralized DNS remains the hard part.
  • Some ask why not just share magnet links directly instead of going through an intermediary website.

Comparisons to IPFS, BitTorrent, and prior projects

  • IPFS is criticized for reliance on gateways and lack of built‑in peer health mechanisms, and for difficult problems around illegal content.
  • BitTorrent is praised as “just works,” especially for large files and Linux distros; mutable torrents and DHT-based search are mentioned as relevant building blocks.
  • WebTorrent is seen as a clever idea that never really took off: few stable WebRTC trackers, blocked or crippled by browser/WebRTC limitations.
  • ZeroNet is cited as a once-performant decentralized web system that is now effectively abandoned.

Technical limitations and performance

  • Many users report that the demo sites never load, or stay stuck on “connecting to peers,” implying either tracker or seeding issues.
  • People stress that >5 seconds to load a page makes it unusable for mainstream web, even if old-timers reminisce about 90s loading times.
  • WebRTC/STUN/NAT traversal are identified as major obstacles to browser-based P2P; ideas include hybrid DHTs (WebRTC + HTTP/WebSocket) and new direct-socket APIs.
  • Suggestions include federated caching servers to improve persistence and offload popular content.

Moderation and legal concerns

  • Serving user-uploaded video at scale is seen as especially risky from a moderation and legal standpoint, worsened by anonymity.
  • Others counter that, since content is shared peer-to-peer among friends and discovery is limited, users largely control what they see.
  • Similar concerns are recalled from IPFS; smaller sites are seen as having smaller risk footprints.

UX, trust, and “AI slop” aesthetics

  • Some appreciate the concept but find the UX confusing and non‑“just works.”
  • The site’s visual style, heavy emoji use, and “vibe-coded” feel lead several to suspect AI‑generated boilerplate and to distrust the seriousness of the project.

Security, sanitization, and addressing

  • The project’s DOMPurify-based HTML sanitization plus iframe sandboxing is noted; some argue sandboxing alone might suffice and that stripping all JS is too aggressive.
  • There are ideas about using the torrent hash (possibly embedding a public key) in subdomains to leverage the browser’s same-origin policy.
  • A few ask whether sanitization can be disabled for serving static sites unchanged.

Related experiments and future directions

  • Commenters share related work: WebTorrent browser extensions, WebRTC Gnutella-like networks, P2P GPU compute in the browser, and older systems like WASTE.
  • One person describes plans for a more robust “Peerweb” platform with anti-abuse mechanisms, automatic asset prioritization, CDN failover, and more polished UX.
  • There is speculation about micropayment incentives for faster service and even decentralized AI infrastructure, with skepticism that economics and complexity have so far stalled a truly distributed web.