3

There are answers related to the number of confirmations is considered secure in Ethereum.

@Waqar Lim

To sum up, it always depends on your application and the attached value to consider which number of confirmations is considered to be secure.

How is it considered for Ethereum PoA network? Do we need to wait for 6 blocks confirmations considered to be secure for Ethereum PoA network?

Is there any possibility to have a fork?

alper
  • 8,395
  • 11
  • 63
  • 152
maroodb
  • 1,111
  • 1
  • 10
  • 32

3 Answers3

1

I guess that depends on your network layout and your level of trust. If it's a big and distributed network there is bound to be some latency and you will get sub-chains. But if it's small and trusted you will get a lot less sub-chains. I guess PoA chains are typically quite small (maybe?).

The amount of confirmations you want to wait for is up to you. In the Ethereum mainnet some exchanges require 10 confirmation. In a much smaller network the amount should probably be lower, but it's all up to you. The more confirmations you wait for the more secure the transaction is.

Lauri Peltonen
  • 29,391
  • 3
  • 20
  • 57
1

In addition to the above, it depends on which flavour of PoA you are running.

Clique has determinstic finality and is works off GHOST (Greediest Observed Heavy Subtree). So depending on your trust model, you might want to wait a couple of confirmations.

Aura on the other hands provides finality after a majority vote i.e greater than n/2 validators have voted on the block, so ideally it would depend on the number of validators and your stepDuration

References

Community
  • 1
0xsegfault
  • 1,260
  • 9
  • 24
  • How is it handle on Rinkeby, which is large network? @Samuel Dare – alper Jul 29 '19 at 21:21
  • Although my answer naively assumed a private chain, Rinkeby is a proof of authority testnet based on Clique (EIP Here: https://github.com/ethereum/EIPs/issues/225). So you are implictly trusting the validators already and shouldnt have to worry too much about validators on Rinkeby acting malicious ; there are not incentives and the validators are trusted – 0xsegfault Jul 29 '19 at 21:27
  • 1
    Since the validators are trusted; I think 0 block confirmation should be reasonable. I think that could be the case on the private PoA created by Clique as well since the validators are selected via voting and should be trusted as well in my opinion. @Samuel Dare – alper Jul 29 '19 at 21:32
  • Yes for Byzantine Faults.. but you could still have omissions and crashes .. i.e. hard ware failures.. network paritiions etc...so your trust is bounded.. nevertheless I wouldnt care so much for Rinkeby seeing as its a testnet. Depending on the economic value at risk you might want to opt for algorithms with instant finality eg. IBFT2.0 and Honey Badger – 0xsegfault Jul 29 '19 at 21:36
  • I wonder how https://poa.network or even eos handles it since its based on a economic value. Since their block-time is much less than the ethereum-mannet, the confirmation block number should be even more for those networks than the ethereum mainnet. @Samuel Dare – alper Jul 29 '19 at 21:46
  • 1
    EoS is another kettle of fish and I am neither qualified nor have any desire to comment on . PoS_DAO uses a variant of Aura so its finality rules more or less apply there. They are also rapidly working on Honey Badger which i think is a far superior algo. The confirmation blocks number is based on how reversible transactions are in the consensus algorithms. In aura once you have greater than n/2 validators signed , the block is finalised so number of confirmations is related to the number of validators.. not neccesarily block confirmation rounds – 0xsegfault Jul 29 '19 at 21:50
  • I think In aura once you have greater than n/2 validators signed , the block is finalised => this is the case in Clique as well right? After the block is finalised there will be no additinal block confirmation. – alper Jul 29 '19 at 21:58
  • 1
    check out this paper for my rational behind clique : http://ceur-ws.org/Vol-2058/paper-06.pdf – 0xsegfault Jul 29 '19 at 22:05
0

Who can join this PoA network? Do you trust them all 100%?

If so, you do not need to wait for 6 confirmations.

Matt Swezey
  • 1,246
  • 6
  • 13