On 12/04/2016 11:13 PM, David Barrett wrote: > My sense is p2p tech is on the downswing, but not forever.
Which of course begs the question, what is needed for a p2p resurgence. During the early rise of P2P (bittorrent, skype, and various file sharing apps) the vast majority of consumer devices were AC powered. So things like meshs, DHTs, and related technologies that have frequent and mostly random network connections mostly worked. Sure IP Masq/NAT is nasty, but it was mostly worked around. Then most devices (percentage wise) went mobile. Suddenly receiving a packet a few times a second is unacceptable (both because of bandwidth and battery usage). Simultaneously users started expecting more performance. Quick launching for streaming video, tiny latencies for messaging (or even presence messages like "I'm typing"), and high bandwidth (1080P or 4k streaming). Some claim Microsoft is just evil and spying, but I believe you can justify their move from p2p (self promoting supernodes and normal nodes) to a centralized infrastructure makes sense when a large fraction of your nodes are phones, tablets, or laptops. I believe what's needed to fix p2p and allow p2p to match more traditional providers like dropbox, box.com, skype, gmail.com, etc. is a 2 tiered approach. A supernode tier for AC connected widgets with an "always on" network connection. IPv6 is a plus, but even a PI or one of the arm based "plug" computers would be fine. Said supernodes could: * store encrypted blobs on your behalf * earn good will from random peers by trading storage, bandwidth, or proof of work. * queue incoming messages for you (but replicate on peers as well) * have at least 64 GB of storage ($20 of microSD is fine) * could easily handle DHT membership traffic, keep an up to date DHT * designed to be online 99% or more of the time * should cost $100 * be able to push data to a client. * potentially store/forward encrypted blobs to help with anonymity Then regular clients: * could check in with a supernode (preferably yours, but if need be others) * could collect on the good will of a supernode you run * would allow collecting RSS feeds, email, IM, etc * would be a DHT client (for service discovery), but not a full node * generally spends most of it's time off, asleep, or in low power mode That way the hard parts of P2P (things like power efficiency, handling when sender and receiver are offline, and latency) would be easier. Seems like what we need are some well defined APIs, and a popular $50-$100 piece of hardware that allows users to join by paying, plugging, and typing in nothing much more complicated than a username and a password. Thoughts? _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
