6

Lets say a node has a very high delay (ping/pong) between its own network and the network on which most of the other nodes are, say about 13 minutes, will it be able to be part of the main blockchain (and will it be able do act as any other nodes, not just reading the chain state but also commit transactions) or will it have to connect on a blockchain with pairs that are close to him (a few hundred milliseconds) and have a bridge between its chain and the far far away one ?

Thanks.

eth
  • 85,679
  • 53
  • 285
  • 406
Nicolas Massart
  • 6,783
  • 2
  • 29
  • 63

1 Answers1

4

Not with the current protocol. Such a feature would require a so-called delay/disruption tolerant networking, which is seriously out of scope of the project for now. Such a protocol has a lot of completely unique challenges and there's no real point in trying to even design such a thing until a realistic use case arises. So unless NASA opens a new challenge to deploy Ethereum on the ISS, I don't think we're going to see much need for it.

So what would be the maximum delay ? The downloader in Geth has a 3.5 second timeout for retrieving a batch of headers from a remote node. So if we consider the node needs to look up the headers too, about 1.5 second one way latency would be the cap.

Nicolas Massart
  • 6,783
  • 2
  • 29
  • 63
Péter Szilágyi
  • 10,436
  • 39
  • 42
  • So, to complete your answer, would you have any idea of what would be the highest acceptable delay with the current protocols? – Nicolas Massart May 06 '16 at 15:17
  • 1
    The downloader in Geth has a 3.5 second timeout for retrieving a batch of headers from a remote node. So if we consider the node needs to look up the headers too, about 1.5 second one way latency would be the cap. This will probably be relaxed a bit in Geth 1.5 with the rollout of concurrent header downloads, but we didn't want to stall syncs due to very slow peers. – Péter Szilágyi May 06 '16 at 15:39
  • 1
    Theoretically latency to Moon is around 1 second. So it's doable. it's Mars that's completely out of question, with 20 minutes of latency. –  May 06 '16 at 18:51
  • Moon is too close, I'm not interested ;) But what if we spread space relays that run nodes all along the path to mars on order to maintain 1s delay between each node ? Would it change something ? – Nicolas Massart May 06 '16 at 21:48
  • 1
    It would make more sense to use a custom protocol that broadcasts (suitably erasure coded) blocks. It'd certainly be cheaper than building many concentric rings of satellites in solar orbit! – Nick Johnson May 06 '16 at 22:31