Re:Meta torrented & seeded 81.7 TB dataset containing copyri...
by lysace on Friday February 07, 2025 @10:51AM
(#42974056)
Afaik, the trick was to stream (via http, I assume) the first few hundred kilobytes or so from fast/expensive servers and then try to p2p the rest in some clever order. I guess seeking also triggered the fast/expensive path for a while.
Another interesting note, in the early days of spotify, the app would saturate your upload bandwidth while using it. Given their close ties to utorrent, I always assumed that's how they were affording the bandwidth as well.
Pretty brilliant way to bootstrap I guess; they didn't have to pay for bandwidth or content until they already had contracts in place