5 ms·

Yet another reason why Proof of Stake is destined to outcompete Proof of Work. ;)

garmaine 8y ago
Proof of Stake is a perpetual motion machine. You can't safely tie consensus to something (stake) which is determined by consensus. That's circular reasoning and ungrounded.
- drb91 8y ago
  What would failure look like?
  - garmaine 8y ago
    A system that has effectively bifurcated with conflicting chain tips caused by rewriting history with new nodes unable to tell which is original, or a fully centralized stake-mining where a subset of stake voters game the system to unfairly select themselves out of proportion of their stake weight.
    - koheripbal 8y ago
      Is that reeealy so different from a PoW fork?
      - garmaine 8y ago
        A proof-of-work fork is an unstable situation, resolved when more work is applied to one side vs the other, which inevitably happens without active intervention. A proof-of-stake fork is permanent--someone new to the network has no built-in mechanism for deciding which side to prefer: they both seem equally valid. A properly constructed proof-of-work mechanism has no built-in mechanism for a malicious minority hash power to take control of the protocol, shutting out the higher hash rate majority from mining blocks.
- jki275 8y ago
  And yet we have multiple Proof of stake coins that seem to work just fine.
  - garmaine 8y ago
    Until they don't. There have been proof-of-stake failures before, and there will be more. Every proof-of-stake system is either (a) known to be broken, or (b) too complicated to analyze to show how it is broken. As the exploitable order depth in category (b) increases, it becomes worthwhile to invest the hours, days, weeks, months it takes to develop attacks against these Rube Goldberg contraptions and they get moved to category (a).
    - jki275 8y ago
      That's an assertion that I haven't seen any data to back. I don't believe either (a) or (b) applies to either xtz or to ada, though ada hasn't yet deployed POS. I can well believe that there are broken POS implementations, there have been a lot of broken implementations of a lot of things in the space -- but claiming every system is known to be broken or cannot be analyzed is a very bold claim to make.
      - garmaine 8y ago
        How would you expect a statement that certain mechanisms have become too complex for review to be backed by data? What observations could I convey to you over a HN comment other than “go look for yourself”? For a write up of the general, unavoidable problems of proof of stake, see Andrew Poelstra’s proof of stake FAQ.
        jki275 8y ago
        I gave you two of the best known examples of proof of stake that there are. The Tezos team has done it and it's been running for about a year now. The Cardano team is doing it, and I'll link their paper below. As far as Poelstra, that's a rather old document. A great deal of research has been done in the field since. Cardano's paper: https://eprint.iacr.org/2016/889.pdf https://eprint.iacr.org/2016/889.pdf

Proof of Stake is a perpetual motion machine. You can't safely tie consensus to something (stake) which is determined by consensus. That's circular reasoning and ungrounded.

drb91 8y ago
What would failure look like?
- garmaine 8y ago
  A system that has effectively bifurcated with conflicting chain tips caused by rewriting history with new nodes unable to tell which is original, or a fully centralized stake-mining where a subset of stake voters game the system to unfairly select themselves out of proportion of their stake weight.
  - koheripbal 8y ago
    Is that reeealy so different from a PoW fork?
    - garmaine 8y ago
      A proof-of-work fork is an unstable situation, resolved when more work is applied to one side vs the other, which inevitably happens without active intervention. A proof-of-stake fork is permanent--someone new to the network has no built-in mechanism for deciding which side to prefer: they both seem equally valid. A properly constructed proof-of-work mechanism has no built-in mechanism for a malicious minority hash power to take control of the protocol, shutting out the higher hash rate majority from mining blocks.
jki275 8y ago
And yet we have multiple Proof of stake coins that seem to work just fine.
- garmaine 8y ago
  Until they don't. There have been proof-of-stake failures before, and there will be more. Every proof-of-stake system is either (a) known to be broken, or (b) too complicated to analyze to show how it is broken. As the exploitable order depth in category (b) increases, it becomes worthwhile to invest the hours, days, weeks, months it takes to develop attacks against these Rube Goldberg contraptions and they get moved to category (a).
  - jki275 8y ago
    That's an assertion that I haven't seen any data to back. I don't believe either (a) or (b) applies to either xtz or to ada, though ada hasn't yet deployed POS. I can well believe that there are broken POS implementations, there have been a lot of broken implementations of a lot of things in the space -- but claiming every system is known to be broken or cannot be analyzed is a very bold claim to make.
    - garmaine 8y ago
      How would you expect a statement that certain mechanisms have become too complex for review to be backed by data? What observations could I convey to you over a HN comment other than “go look for yourself”? For a write up of the general, unavoidable problems of proof of stake, see Andrew Poelstra’s proof of stake FAQ.
      - jki275 8y ago
        I gave you two of the best known examples of proof of stake that there are. The Tezos team has done it and it's been running for about a year now. The Cardano team is doing it, and I'll link their paper below. As far as Poelstra, that's a rather old document. A great deal of research has been done in the field since. Cardano's paper: https://eprint.iacr.org/2016/889.pdf https://eprint.iacr.org/2016/889.pdf