Conversation

Notices

1. Embed this notice
   deutrino (deutrino@mstdn.io)'s status on Monday, 14-Oct-2024 03:59:20 JST deutrino

   does anybody else who's fascinated with #p2p #decentralized #DHT apps remember some ancient p2p network - maybe ed2k, that era - which would fall back to scanning all of ipv4 space to find other nodes if it couldn't connect to any other nodes or bootstrap servers at startup?

   I read a paper on this in the 2000s. there was a degree of analysis of how long it'd take the network to re-establish itself given n disconnected nodes at the start. I'm looking for that paper/article.

   boosts appreciated!

   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Monday, 14-Oct-2024 03:59:19 JST Sick Sun

     in reply to
     @deutrino That sounds like something a particular GNUtella client might do but not in the protocol itself.
   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Monday, 14-Oct-2024 04:00:46 JST Sick Sun

     in reply to
     @deutrino I am betting yu will find that it's a client-specific innovation of just one program.
   • Embed this notice
     deutrino (deutrino@mstdn.io)'s status on Monday, 14-Oct-2024 04:00:47 JST deutrino

     in reply to

     • Sick Sun

     @sun eDonkey2000 and Gnutella are my two main suspects, based on very hazy memories from the mid-2000s of reading deep dives on P2P implementations

   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Monday, 14-Oct-2024 04:03:44 JST Sick Sun

     in reply to
     @deutrino the reason I suspect gnutella is it did a lot of things for network discovery and nat punching in later versions so that seems like the kind of thing it would do. it was popular inside college universities so it wuld make sense to scan local network
   • Embed this notice
     deutrino (deutrino@mstdn.io)'s status on Monday, 14-Oct-2024 04:03:45 JST deutrino

     in reply to

     • Sick Sun

     @sun noooo :( but maybe.

     last I knew, ed2k turned into a zombie p2p network for botnet C&C, so maybe more than one client :)

     Sick Sun likes this.
   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Monday, 14-Oct-2024 04:37:19 JST Sick Sun

     in reply to

     • Δж3
     @deutrino @ax3 the paper doesn't say it (I don't think anyone knew it yet) but you can bifurcate a chord network by rapidly adding and removing evil nodes.
   • Embed this notice
     deutrino (deutrino@mstdn.io)'s status on Monday, 14-Oct-2024 04:37:20 JST deutrino

     in reply to

     • Sick Sun
     • Δж3

     @ax3 @sun this looks more like object discovery than peer discovery to me after skimming a couple pages? but, I may be outpacing my level of caffeine intake for the time of day here 😆

     what I did see in the skim is very interesting however:

     > Three features that distinguish Chord from many other peer-to-peer lookup protocols are its simplicity, provable correctness, and provable performance. Chord is simple, routing a key through a sequence of O(log n) other nodes toward the destination.

     #p2p

   • Embed this notice
     Δж3 (ax3@wizard.casa)'s status on Monday, 14-Oct-2024 04:37:26 JST Δж3

     in reply to

     • Sick Sun

     @deutrino @sun this may be whitepaper you're referring to https://pdos.csail.mit.edu/papers/chord:sigcomm01/chord_sigcomm.pdf there may be others but this was a good whitepaper to read back then.

   • Embed this notice
     Δж3 (ax3@wizard.casa)'s status on Monday, 14-Oct-2024 04:57:19 JST Δж3

     in reply to

     • Sick Sun

     @sun @deutrino

     > the paper doesn't say it (I don't think anyone knew it yet) but you can bifurcate a chord network by rapidly adding and removing evil nodes. I think this is why everyone uses Kademlia now

     you're right. it's a downside of chord. but would be interesting to simulate.

     Sick Sun likes this.