On Fri, Dec 27, 2013 at 4:28 AM, Bill Broadley <bill@broadley.org> wrote:

On 12/24/2013 01:20 AM, grarpamp wrote:

...

grarpamp... Bittorrent is already in the 100m node range.

Numbers I've seen show 8-10M for users in the DHT at any one time. If it's actually 100M all the better.

I think if you load Vuze with the mlDHT plugin you'll often see 150m users online.

...

That's not enough. This needs to replace every possible messaging user on the planet over the duration of their actiive lifetime. That's at least a couple billion nodes. Don't forget, you can always use disk to cache things.

Considering the DHT already scales to 2^23 peers, what causes you to think the next 2^7th is going to cause problems? Especially when router table and traffic increases with the log(peers)?

Current bittorrent clients often have 20 bins in use, tracking about 160 peers per 15 minutes. That would only change toe 27 bins (216 hosts) for 1 billion peers. Seems workable to me. Did you have some specific concerns?

If a DHT can scale to say 10B nodes while performing lookup on an unknown key in say a minute or less [1], that sounds like a great start. Are there such designs in effect? The concern is that we don't appear to have any decentralized p2p messaging network today that is anywhere near that large. When you ask the current big ones (BT, Tor, I2P, cjdns, etc) they don't seem to have a scale solution, be they filesharing, transport, messaging, etc. [1] Perhaps reasonable latency for delivery of mail across an anonymous transport (aka: circuit setup time, uncached) may be a few minutes or so.