4 ms·

The Architecture of a Cryptocurrency

51 points by kushti 11y ago

nickodell 11y ago
The point about ethereum's internal state is good, but consider what could happen if two nodes disagree about the state of a contract: You'd wind up in a situation where different nodes could potentially authorize different payments.
- kushti 11y ago
  That's true. My fear is the problem is possible though in some form with software updates / fixes seems to be unavoidable.
- yarrel 11y ago
  Totally. That's why, as with Bitcoin, you wait before accepting the state of the network.
sktrdie 11y ago
I don't think there has been any empirical evidence regarding the idea that proof-of-stake method can reach consensus, so I'm not sure about the quality of this document. It is sort of implied, but in reality one can easily fake n chains and there would be no way for a peer to independently figure out which one is valid. In proof-of-work a peer can always rely on the idea that the chain with the most work is the valid chain. Proof-of-stake is an interesting idea, but until I see empirical evidence showing that it actually works, I remain skeptical. Here's a good read on the subject: https://download.wpsoftware.net/bitcoin/pos.pdf https://download.wpsoftware.net/bitcoin/pos.pdf
- kushti 11y ago
  Author is here! Here's my feedback and discussion with Poelstra https://www.reddit.com/r/Bitcoin/comments/2zpmlj/expanded_rewrite_of_distributed_consensus_from/cplj4ug https://www.reddit.com/r/Bitcoin/comments/2zpmlj/expanded_re... . There's no any formal proof of consensus (im)-possibility around. But we have some cryptocurrencies successfully working for years already(despite promises to hack them easily), also there're some simulations around, e.g. ours https://github.com/ConsensusResearch/ForgingSimulation https://github.com/ConsensusResearch/ForgingSimulation and https://github.com/ConsensusResearch/MultiBranch https://github.com/ConsensusResearch/MultiBranch (corresponding articles https://github.com/ConsensusResearch/articles-papers https://github.com/ConsensusResearch/articles-papers ). I hope more formal framework will be established someday, and would be happy to help with that(I can contribute with Coq code etc).
  - sktrdie 11y ago
    There's a proof in Bitcoin's paper that shows the probability of someone catching up with the main chain. That's actual real data that we can measure. Of course nothing's 100% true in the real-world, but with PoW you get very real numbers on the probability of someone catching up, which decreases drastically after each confirmation. On the other hand with PoS I still haven't seen any data whatsoever. I asked a simple question: since there's no cost for someone to simulate a chain, what's stopping peers from doing this? I know there are checkpoints and things like Nxt don't allow deep reorgs, but for new users joining the network it means you need to point them to trusted parties that can give them this information: they can't trust the chain independently.
    - kushti 11y ago
      In the first place, by using modern ASIC it's possible to draw a chain from the genesis having no difficulty changes(by setting proper timestamps), and it will be longer than network's probably as latter has delays more than 10 mins pretty often. So Bitcoin has checkpoint as well ( https://github.com/bitcoin/bitcoin/blob/86cfd23f68367af072500b1758a4c446cdd36e74/src/chainparams.cpp https://github.com/bitcoin/bitcoin/blob/86cfd23f68367af07250... ). In the second place a chance of drawing a long chain suffix with cumulative difficulty better than network's is negligible. Short-range attacks are possible, but they are not a huge problem. Please read the forum post with some results got with simulations https://bitcointalk.org/index.php?topic=897488 https://bitcointalk.org/index.php?topic=897488
      - sktrdie 11y ago
        What do you mean? If I look at the latest block hash and difficulty I can actually measure how much hashing power went into finding such block (and it's a lot). How would you simulate that with timestamps?
        kushti 11y ago
        You can generate blockchain since genesis setting timestamps every 10 mins so no any difficulty change. And that chain drawn to the moment will be longer than network's.
l_m_g_ 11y ago
It's nice to see the Tezos paper becoming influential, but attribution would have been extra nice.
- infruset 11y ago
  Hi, I've been trying to find a way to contact you about the Tezos paper, which I found very interesting. Are you working on an implementation? Is there any way to participate in the process?
  - l_m_g_ 11y ago
    There's a working beta which is being tested for scalability at the moment. Best way to reach is twitter DM.
    - infruset 11y ago
      Great! I just saw your reply, I don't know your twitter handle (it doesn't seem to be @l_m_g_ or @tezos).
- kushti 11y ago
  Right, the idea of the three protocols was initially explicitly stated in the Tezos position paper http://tezos.com/position_paper.pdf http://tezos.com/position_paper.pdf. Initial draft had a reference to it even but then it was deleted as "References" section had only one reference :) Now "Acknowledgements" section has been added. P.S. How's Tezos going?
nickpsecurity 11y ago
I think the focus has been too narrow because of Bitcoin's success and the document reflects it. The first cryptocurrency, Chaum's eCash, wouldn't meet these criteria. And reputable banks using fiat currency are still more trustworthy for storing or transferring money than Bitcoin. Our mental framework for cryptocurrency should include the possibility of such centralized models as they're easier, more efficient, and can be quite trustworthy with right jurisdictions + charters + TLA's/contracts. Them plus something like Open Transactions. Note: A side advantage is that the bank's investments and fees can fund the system. Another is possible acceptance by regulators if it's (a) easy to regulate and (b) tied to their currency to support rather than compete with it.

51 points by kushti 11y ago

The point about ethereum's internal state is good, but consider what could happen if two nodes disagree about the state of a contract: You'd wind up in a situation where different nodes could potentially authorize different payments.

kushti 11y ago
That's true. My fear is the problem is possible though in some form with software updates / fixes seems to be unavoidable.
yarrel 11y ago
Totally. That's why, as with Bitcoin, you wait before accepting the state of the network.

sktrdie 11y ago

I don't think there has been any empirical evidence regarding the idea that proof-of-stake method can reach consensus, so I'm not sure about the quality of this document. It is sort of implied, but in reality one can easily fake n chains and there would be no way for a peer to independently figure out which one is valid. In proof-of-work a peer can always rely on the idea that the chain with the most work is the valid chain. Proof-of-stake is an interesting idea, but until I see empirical evidence showing that it actually works, I remain skeptical. Here's a good read on the subject: https://download.wpsoftware.net/bitcoin/pos.pdf https://download.wpsoftware.net/bitcoin/pos.pdf

kushti 11y ago
Author is here! Here's my feedback and discussion with Poelstra https://www.reddit.com/r/Bitcoin/comments/2zpmlj/expanded_rewrite_of_distributed_consensus_from/cplj4ug https://www.reddit.com/r/Bitcoin/comments/2zpmlj/expanded_re... . There's no any formal proof of consensus (im)-possibility around. But we have some cryptocurrencies successfully working for years already(despite promises to hack them easily), also there're some simulations around, e.g. ours https://github.com/ConsensusResearch/ForgingSimulation https://github.com/ConsensusResearch/ForgingSimulation and https://github.com/ConsensusResearch/MultiBranch https://github.com/ConsensusResearch/MultiBranch (corresponding articles https://github.com/ConsensusResearch/articles-papers https://github.com/ConsensusResearch/articles-papers ). I hope more formal framework will be established someday, and would be happy to help with that(I can contribute with Coq code etc).
- sktrdie 11y ago
  There's a proof in Bitcoin's paper that shows the probability of someone catching up with the main chain. That's actual real data that we can measure. Of course nothing's 100% true in the real-world, but with PoW you get very real numbers on the probability of someone catching up, which decreases drastically after each confirmation. On the other hand with PoS I still haven't seen any data whatsoever. I asked a simple question: since there's no cost for someone to simulate a chain, what's stopping peers from doing this? I know there are checkpoints and things like Nxt don't allow deep reorgs, but for new users joining the network it means you need to point them to trusted parties that can give them this information: they can't trust the chain independently.
  - kushti 11y ago
    In the first place, by using modern ASIC it's possible to draw a chain from the genesis having no difficulty changes(by setting proper timestamps), and it will be longer than network's probably as latter has delays more than 10 mins pretty often. So Bitcoin has checkpoint as well ( https://github.com/bitcoin/bitcoin/blob/86cfd23f68367af072500b1758a4c446cdd36e74/src/chainparams.cpp https://github.com/bitcoin/bitcoin/blob/86cfd23f68367af07250... ). In the second place a chance of drawing a long chain suffix with cumulative difficulty better than network's is negligible. Short-range attacks are possible, but they are not a huge problem. Please read the forum post with some results got with simulations https://bitcointalk.org/index.php?topic=897488 https://bitcointalk.org/index.php?topic=897488
    - sktrdie 11y ago
      What do you mean? If I look at the latest block hash and difficulty I can actually measure how much hashing power went into finding such block (and it's a lot). How would you simulate that with timestamps?
      - kushti 11y ago
        You can generate blockchain since genesis setting timestamps every 10 mins so no any difficulty change. And that chain drawn to the moment will be longer than network's.

l_m_g_ 11y ago

It's nice to see the Tezos paper becoming influential, but attribution would have been extra nice.

infruset 11y ago
Hi, I've been trying to find a way to contact you about the Tezos paper, which I found very interesting. Are you working on an implementation? Is there any way to participate in the process?
- l_m_g_ 11y ago
  There's a working beta which is being tested for scalability at the moment. Best way to reach is twitter DM.
  - infruset 11y ago
    Great! I just saw your reply, I don't know your twitter handle (it doesn't seem to be @l_m_g_ or @tezos).
kushti 11y ago
Right, the idea of the three protocols was initially explicitly stated in the Tezos position paper http://tezos.com/position_paper.pdf http://tezos.com/position_paper.pdf. Initial draft had a reference to it even but then it was deleted as "References" section had only one reference :) Now "Acknowledgements" section has been added. P.S. How's Tezos going?

nickpsecurity 11y ago

I think the focus has been too narrow because of Bitcoin's success and the document reflects it. The first cryptocurrency, Chaum's eCash, wouldn't meet these criteria. And reputable banks using fiat currency are still more trustworthy for storing or transferring money than Bitcoin. Our mental framework for cryptocurrency should include the possibility of such centralized models as they're easier, more efficient, and can be quite trustworthy with right jurisdictions + charters + TLA's/contracts. Them plus something like Open Transactions. Note: A side advantage is that the bank's investments and fees can fund the system. Another is possible acceptance by regulators if it's (a) easy to regulate and (b) tied to their currency to support rather than compete with it.