monsterer
Legendary
Offline
Activity: 1008
Merit: 1007
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 18, 2016, 09:18:29 PM |
|
Why does temporarily halting the network = drive the price down to 0? Once it is halted, people could simply take action and vote out the evil witnesses. While similar to >51% control in PoW the outcome for halting a DPoS network appears much less appealing...
How are people going to vote when the network is halted? Votes are transactions.
|
|
|
|
TPTB_need_war (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 20, 2016, 12:28:50 PM Last edit: February 20, 2016, 12:41:48 PM by TPTB_need_war |
|
Retarded egos versus rationality... It seems what ever design you contemplate employing a fee can always be gamed by an attacker so that he pays no cost.
The attacker can make his own DE and receive the fees. So then he attacks the other DEs (which are honest) 100% of the time but doesn't attack his own DE (or attacks he own less frequently) thus all users migrate to his DE. Thus he is losing much less in attack fees than he is gaining from fees.
Since more than one attacker can do this to each other, no one will use DE. P.S. I knew that there would be a leak in fees that would maintain my conclusion that DE is fundamentally impossible. Sorry when I come to these sort of generative essence conclusions they stick.
By this logic no cryptocoin would exist. Instead of all out war all the time, maybe some sort of truce would be made? Why would everyone running a DE want to spend all that time attacking the other DE's just to get attacked back and then nobody's works. Seems like MAD. I notice your attack scenarios are getting more and more convoluted as now the attacker needs to run a DE themselves to make it even a potential issue. So things are improving. It also might be possible to make it so the attacker wont be able to get their fee back, but no script for that yet. However, you must have missed my post where it is possible to identify the attacker. It would be after the fact, but that allows a reputation system to be created and as long as people are careful about filling orders from newbies, there wont be much problems. James P.S. Using your logic, no central exchange would exist as they would all be Ddosing each other all the time. You dont need to pay exchange fees to Ddos an website. DDoS is not an analogy to jamming. Remember the party that goes first in your protocol locks up his funds (e.g. UXTO) for the period of the timeout refund (which is orders-of-magnitude greater delay than processing a DDoS packet and the attackers delay is asymmetrically orders-of-magnitude smaller) while waiting for the counter party to reciprocate. In bandwidth DDoS, the attacker must have enough bots to saturate the bandwidth of the victim, which is a significant resource. In malformed requests DDoS, the cost to the victim of verifying needs to the symmetric to the cost of the attacker to sign the request. I wrote a white paper about how DDoS can be ameliorated with proper design in the crypto currency setting. Please make sure you understand the relevance of the word 'asymmetry' of costs for the attacker versus the victim w.r.t. DoS. Seems that running a DE in your original fee-based scheme (which I allege is broken) is quite simple as it only involves offering a payto address. Sorry to be a pain, but really you should thank those who do peer review and help you from wasting your time. I'd really appreciate a more cordial and rational response. Getting irrationally upset at those who try to help you, is what the retards over in the Altcoin Discussion forum do, and I know you are too mature and do not possess that behavior. Please remember in the past when you made similar arguments about why proof-of-stake was not attackable, which ended up being wrong as evidenced by the current top-down governance of Nxt and Bitshares. Those of us who take the time to analyze and point out egregious flaws do it because we want correct solutions, not because we are haters. If those we help, hate back on us, the entire discussion forums descends into stupid noise. Please even if you have vested interests, still try to remain fair and rational. This is a scientific discussion, and let's not conflate it with other matters such as any investment you've already made in DE. I can't do my work in a rational, scientific way if I have to worry about each person's ego and vested interests that I might offend by writing rationally. I didn't see where you could identify the attacker in your protocol without incurring any possibility of bailing out and jamming. Perhaps you can explain that more succinctly (without all the cut & choose employing Bitcoin op codes since I don't comprehend Bitcoin op codes thus I can't understand the cut & choose in the way you guys have explained it thus far, although I think I understand the concept which is to probabilistically reveal the preimage). Any way, I was thinking about this more because I know you are invested in DE and I really want to help you find a solution. Of course I don't feel good about the situation where you are invested and it is fundamentally flawed. It is not like I get some great joy in finding flaws. I get great joy in finding solutions that don't have flaws. I can't help it that there is so much flawed work being done in the crypto currency realm. It is the nature of the beast, because crypto logic is complex and there are a lot hair-brained ideas being promulgated. For example, Blockstream's side-chains reduce the security to the weakest chain due to chain reorganizations and the fact that proof-of-work is probabilistic and thus never final (e.g. lie in wait long-con attack). Blockstream is even investing so much effort to push Segregated Witness which enables them to version the block chain with soft forks, thus it looks like a trojan horse so they can get the changes they want to enable this broken concept of side-chains. Speak these rational truths and many people get pissed off and posts get removed by the moderator. Any way, I think I have thought of a solution for DE. The key is to identify the attacker immediately so that all decentralized parties can apply my upthread blocking "Coin Days Destroyed" suggestion. The "Coin Days Destroyed" becomes the reference point that is signed by the owner of the resource, which thus apparently escapes from my generative essence conceptualization of the problem set. So change to the protocol is the provider of the hash sends to the trade's counter party to sign it (hashed with the other party's UXTO address) so the counter party's UXTO can be identified. Then the hash provider (the potential jamming victim) posts this information in a timed refundable transaction to the block chain (spending to the payee contingent on releasing the hash). If the attacker never posts the reciprocal transaction on the other block chain, this enables anyone to identify that attacker and apply the Coin Days Destroyed filtering that I proposed upthread. Note this eliminates the need for any fee. But I assume you can find some justification for a fee, such as perhaps keeping your source code for the DE app closed source and/or offering a centralized fee structure for matching orders, limit orders, etc.. You won't be able to steal funds, which afaik is the most significant advantage of DE over CE.
|
|
|
|
monsterer
Legendary
Offline
Activity: 1008
Merit: 1007
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 20, 2016, 01:52:45 PM |
|
Bringing this topic back on track, I'd like to discuss what I believe is the primary challenge in creating a decentralised, trustless, p2p cryptocurrency. Nash equilibrium for non professional miners.
Because power always centralises, if you reward miners, you create a point of centralisation so the apparent solution is not to reward them at all - instead you make it such that every user has to mine their own transaction in order to participate. And on the surface this seems fine but dig a little deeper and the nash-equilibrium is broken because the cheapest, least risky way (PoW cost and orphaning risk) for users to participate in this process is for them to ignore as much of the network as possible within the bounds of the protocol.
For example, (in a blockchain design with the LCR rule) if you require them to include as many mempool transactions as they can see in their blocks, they are incentivised to pretend they see only their own transaction because the probability of their new block being orphaned by double spend increases with every other transaction they include. This will cause the chain to become divergent because the structure of a chain doesn't permit high frequency blocks resulting from this rational behaviour; the LCR will be wildly selecting between candidate chains in an unintentional orphaning war caused by a combination of this rational behaviour, network latency and the chain structure itself.
In a probabilistic DAG design, even if you have only one transaction per block, orphaning risk still persists; submitting your transaction at the tip of the DAG is more risky than submitting it further down the hierarchy (because probability of being in an orphaned branch decreases with the number of children), so users have an incentive to widen the DAG which is the same as saying they agree not to make forward progress. In addition, users are incentivised to reference as few parents as possible, because the more parents they reference, the higher the chance of their transaction referencing a branch which will become orphaned due to a double spend which isn't apparent when they submit their transaction.
Even if you do away with orphans completely and have a deterministically ordered DAG with one transaction per block, the incentive is for users to reference as few parents as possible because every parent they reference will be given ordering priority over them, which is opposed to the timely confirmation incentive. The narrowest possible DAG is always desirable because the strength of the network decreases linearly in the number of partitions, and partitions are merged by referencing parents.
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
|
|
|
|
David Rabahy
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 20, 2016, 02:01:34 PM |
|
Selfish behavior is great until it wrecks the environment.
|
|
|
|
TPTB_need_war (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 20, 2016, 05:33:22 PM |
|
monsterer there are no mempools in my design.
|
|
|
|
monsterer
Legendary
Offline
Activity: 1008
Merit: 1007
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 20, 2016, 05:40:55 PM |
|
monsterer there are no mempools in my design.
Indeed. I wasn't referring to your design above, just running through potential designs using bitcoin as an analogue.
|
|
|
|
sidhujag
Legendary
Offline
Activity: 2044
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 20, 2016, 10:29:26 PM |
|
Bringing this topic back on track, I'd like to discuss what I believe is the primary challenge in creating a decentralised, trustless, p2p cryptocurrency. Nash equilibrium for non professional miners.
Because power always centralises, if you reward miners, you create a point of centralisation so the apparent solution is not to reward them at all - instead you make it such that every user has to mine their own transaction in order to participate. And on the surface this seems fine but dig a little deeper and the nash-equilibrium is broken because the cheapest, least risky way (PoW cost and orphaning risk) for users to participate in this process is for them to ignore as much of the network as possible within the bounds of the protocol.
For example, (in a blockchain design with the LCR rule) if you require them to include as many mempool transactions as they can see in their blocks, they are incentivised to pretend they see only their own transaction because the probability of their new block being orphaned by double spend increases with every other transaction they include. This will cause the chain to become divergent because the structure of a chain doesn't permit high frequency blocks resulting from this rational behaviour; the LCR will be wildly selecting between candidate chains in an unintentional orphaning war caused by a combination of this rational behaviour, network latency and the chain structure itself.
In a probabilistic DAG design, even if you have only one transaction per block, orphaning risk still persists; submitting your transaction at the tip of the DAG is more risky than submitting it further down the hierarchy (because probability of being in an orphaned branch decreases with the number of children), so users have an incentive to widen the DAG which is the same as saying they agree not to make forward progress. In addition, users are incentivised to reference as few parents as possible, because the more parents they reference, the higher the chance of their transaction referencing a branch which will become orphaned due to a double spend which isn't apparent when they submit their transaction.
Even if you do away with orphans completely and have a deterministically ordered DAG with one transaction per block, the incentive is for users to reference as few parents as possible because every parent they reference will be given ordering priority over them, which is opposed to the timely confirmation incentive. The narrowest possible DAG is always desirable because the strength of the network decreases linearly in the number of partitions, and partitions are merged by referencing parents.
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
Nash also said you win as a whole with the team and lose as an individual, therefor you must act in benefit of the entire network (team) to have the most efficient and effective system. This was his breakthrough where adam smith said the opposite.
|
|
|
|
TPTB_need_war (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 21, 2016, 05:53:28 AM |
|
Bringing this topic back on track, I'd like to discuss what I believe is the primary challenge in creating a decentralised, trustless, p2p cryptocurrency. Nash equilibrium for non professional miners.
Because power always centralises, if you reward miners, you create a point of centralisation so the apparent solution is not to reward them at all - instead you make it such that every user has to mine their own transaction in order to participate. And on the surface this seems fine but dig a little deeper and the nash-equilibrium is broken because the cheapest, least risky way (PoW cost and orphaning risk) for users to participate in this process is for them to ignore as much of the network as possible within the bounds of the protocol.
For example, (in a blockchain design with the LCR rule) if you require them to include as many mempool transactions as they can see in their blocks, they are incentivised to pretend they see only their own transaction because the probability of their new block being orphaned by double spend increases with every other transaction they include. This will cause the chain to become divergent because the structure of a chain doesn't permit high frequency blocks resulting from this rational behaviour; the LCR will be wildly selecting between candidate chains in an unintentional orphaning war caused by a combination of this rational behaviour, network latency and the chain structure itself.
In a probabilistic DAG design, even if you have only one transaction per block, orphaning risk still persists; submitting your transaction at the tip of the DAG is more risky than submitting it further down the hierarchy (because probability of being in an orphaned branch decreases with the number of children), so users have an incentive to widen the DAG which is the same as saying they agree not to make forward progress. In addition, users are incentivised to reference as few parents as possible, because the more parents they reference, the higher the chance of their transaction referencing a branch which will become orphaned due to a double spend which isn't apparent when they submit their transaction.
Even if you do away with orphans completely and have a deterministically ordered DAG with one transaction per block, the incentive is for users to reference as few parents as possible because every parent they reference will be given ordering priority over them, which is opposed to the timely confirmation incentive. The narrowest possible DAG is always desirable because the strength of the network decreases linearly in the number of partitions, and partitions are merged by referencing parents.
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
Nash also said you win as a whole with the team and lose as an individual, therefor you must act in benefit of the entire network (team) to have the most efficient and effective system. This was his breakthrough where adam smith said the opposite. Nash equilibrium doesn't apply to society. It applies in some cases. Adam Smith did not say the opposite.
|
|
|
|
sidhujag
Legendary
Offline
Activity: 2044
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 23, 2016, 04:08:25 AM |
|
Bringing this topic back on track, I'd like to discuss what I believe is the primary challenge in creating a decentralised, trustless, p2p cryptocurrency. Nash equilibrium for non professional miners.
Because power always centralises, if you reward miners, you create a point of centralisation so the apparent solution is not to reward them at all - instead you make it such that every user has to mine their own transaction in order to participate. And on the surface this seems fine but dig a little deeper and the nash-equilibrium is broken because the cheapest, least risky way (PoW cost and orphaning risk) for users to participate in this process is for them to ignore as much of the network as possible within the bounds of the protocol.
For example, (in a blockchain design with the LCR rule) if you require them to include as many mempool transactions as they can see in their blocks, they are incentivised to pretend they see only their own transaction because the probability of their new block being orphaned by double spend increases with every other transaction they include. This will cause the chain to become divergent because the structure of a chain doesn't permit high frequency blocks resulting from this rational behaviour; the LCR will be wildly selecting between candidate chains in an unintentional orphaning war caused by a combination of this rational behaviour, network latency and the chain structure itself.
In a probabilistic DAG design, even if you have only one transaction per block, orphaning risk still persists; submitting your transaction at the tip of the DAG is more risky than submitting it further down the hierarchy (because probability of being in an orphaned branch decreases with the number of children), so users have an incentive to widen the DAG which is the same as saying they agree not to make forward progress. In addition, users are incentivised to reference as few parents as possible, because the more parents they reference, the higher the chance of their transaction referencing a branch which will become orphaned due to a double spend which isn't apparent when they submit their transaction.
Even if you do away with orphans completely and have a deterministically ordered DAG with one transaction per block, the incentive is for users to reference as few parents as possible because every parent they reference will be given ordering priority over them, which is opposed to the timely confirmation incentive. The narrowest possible DAG is always desirable because the strength of the network decreases linearly in the number of partitions, and partitions are merged by referencing parents.
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
Nash also said you win as a whole with the team and lose as an individual, therefor you must act in benefit of the entire network (team) to have the most efficient and effective system. This was his breakthrough where adam smith said the opposite. Nash equilibrium doesn't apply to society. It applies in some cases. Adam Smith did not say the opposite. In away he does.. nash argued a better global optima can ne achieved by the cooperation of a network rather than self-interest. Smiths model would achieve a local optima.
|
|
|
|
spartacusrex
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 23, 2016, 09:47:26 AM |
|
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
What about the 'Security' incentive ? People run full nodes because it is more secure and they don't need to trust anyone. They don't get paid. I suppose it could be classed as financial, as their funds are more secure.
|
Life is Code.
|
|
|
monsterer
Legendary
Offline
Activity: 1008
Merit: 1007
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 23, 2016, 11:23:45 AM |
|
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
What about the 'Security' incentive ? People run full nodes because it is more secure and they don't need to trust anyone. They don't get paid. I suppose it could be classed as financial, as their funds are more secure. But non-miners do not contribute to consensus?
|
|
|
|
spartacusrex
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 23, 2016, 11:41:11 AM |
|
But non-miners do not contribute to consensus?
But in this DAG system there are no miners ? Just users building on top of each other. (Unless we class all the users as 'momentary' miners) And yet there is still a consensus building process. I am saying that there is an incentive for the users to act in the best interest of the whole network, even without paying them. They receive security for their funds. And that this is shown in the btc sphere, where people still run full nodes, for no financial reward.
|
Life is Code.
|
|
|
monsterer
Legendary
Offline
Activity: 1008
Merit: 1007
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 23, 2016, 11:48:15 AM |
|
But non-miners do not contribute to consensus?
But in this DAG system there are no miners ? Just users building on top of each other. (Unless we class all the users as 'momentary' miners) And yet there is still a consensus building process. I am saying that there is an incentive for the users to act in the best interest of the whole network, even without paying them. They receive security for their funds. And that this is shown in the btc sphere, where people still run full nodes, for no financial reward. In the DAG systems, users are most definitely miners, yes. I assumed you were talking about bitcoin, because you said 'full node'. Your assumption about security isn't considering what is actually most secure for them - double spends cause all kinds of nasty problems, which I discuss above.
|
|
|
|
spartacusrex
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 23, 2016, 12:14:19 PM |
|
Your assumption about security isn't considering what is actually most secure for them - double spends cause all kinds of nasty problems, which I discuss above.
Yep - I can't see a clean solution to this issue of double spends in the current DAG chain approach. Other than ONLY building on chains that have no double spends in them. So that there is a valid version of history. A user would not build on top of a txn-chain that has inconsistencies and can always go back through history and build on top of the last 'fully' valid txn-chain he finds. Would this simply cause the chain to explode 'width-wise' ?
|
Life is Code.
|
|
|
monsterer
Legendary
Offline
Activity: 1008
Merit: 1007
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 23, 2016, 03:03:30 PM |
|
Would this simply cause the chain to explode 'width-wise' ?
Exactly - this is like saying the best option is to agree not to make forward progress; a disaster for convergence.
|
|
|
|
TPTB_need_war (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 24, 2016, 11:12:26 PM |
|
Bringing this topic back on track, I'd like to discuss what I believe is the primary challenge in creating a decentralised, trustless, p2p cryptocurrency. Nash equilibrium for non professional miners.
Because power always centralises, if you reward miners, you create a point of centralisation so the apparent solution is not to reward them at all - instead you make it such that every user has to mine their own transaction in order to participate. And on the surface this seems fine but dig a little deeper and the nash-equilibrium is broken because the cheapest, least risky way (PoW cost and orphaning risk) for users to participate in this process is for them to ignore as much of the network as possible within the bounds of the protocol.
For example, (in a blockchain design with the LCR rule) if you require them to include as many mempool transactions as they can see in their blocks, they are incentivised to pretend they see only their own transaction because the probability of their new block being orphaned by double spend increases with every other transaction they include. This will cause the chain to become divergent because the structure of a chain doesn't permit high frequency blocks resulting from this rational behaviour; the LCR will be wildly selecting between candidate chains in an unintentional orphaning war caused by a combination of this rational behaviour, network latency and the chain structure itself.
In a probabilistic DAG design, even if you have only one transaction per block, orphaning risk still persists; submitting your transaction at the tip of the DAG is more risky than submitting it further down the hierarchy (because probability of being in an orphaned branch decreases with the number of children), so users have an incentive to widen the DAG which is the same as saying they agree not to make forward progress. In addition, users are incentivised to reference as few parents as possible, because the more parents they reference, the higher the chance of their transaction referencing a branch which will become orphaned due to a double spend which isn't apparent when they submit their transaction.
Even if you do away with orphans completely and have a deterministically ordered DAG with one transaction per block, the incentive is for users to reference as few parents as possible because every parent they reference will be given ordering priority over them, which is opposed to the timely confirmation incentive. The narrowest possible DAG is always desirable because the strength of the network decreases linearly in the number of partitions, and partitions are merged by referencing parents.
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
Nash also said you win as a whole with the team and lose as an individual, therefor you must act in benefit of the entire network (team) to have the most efficient and effective system. This was his breakthrough where adam smith said the opposite. Nash equilibrium doesn't apply to society. It applies in some cases. Adam Smith did not say the opposite. In away he does.. nash argued a better global optima can ne achieved by the cooperation of a network rather than self-interest. Smiths model would achieve a local optima. Again Nash equilibrium only applies where the incentives are so aligned. It doesn't apply generally in society with unbounded entropy. Sorry please pay attention to what I write, so I don't have to repeat myself.
|
|
|
|
TPTB_need_war (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 24, 2016, 11:14:41 PM |
|
Bringing this topic back on track, I'd like to discuss what I believe is the primary challenge in creating a decentralised, trustless, p2p cryptocurrency. Nash equilibrium for non professional miners.
Because power always centralises, if you reward miners, you create a point of centralisation so the apparent solution is not to reward them at all - instead you make it such that every user has to mine their own transaction in order to participate. And on the surface this seems fine but dig a little deeper and the nash-equilibrium is broken because the cheapest, least risky way (PoW cost and orphaning risk) for users to participate in this process is for them to ignore as much of the network as possible within the bounds of the protocol.
For example, (in a blockchain design with the LCR rule) if you require them to include as many mempool transactions as they can see in their blocks, they are incentivised to pretend they see only their own transaction because the probability of their new block being orphaned by double spend increases with every other transaction they include. This will cause the chain to become divergent because the structure of a chain doesn't permit high frequency blocks resulting from this rational behaviour; the LCR will be wildly selecting between candidate chains in an unintentional orphaning war caused by a combination of this rational behaviour, network latency and the chain structure itself.
In a probabilistic DAG design, even if you have only one transaction per block, orphaning risk still persists; submitting your transaction at the tip of the DAG is more risky than submitting it further down the hierarchy (because probability of being in an orphaned branch decreases with the number of children), so users have an incentive to widen the DAG which is the same as saying they agree not to make forward progress. In addition, users are incentivised to reference as few parents as possible, because the more parents they reference, the higher the chance of their transaction referencing a branch which will become orphaned due to a double spend which isn't apparent when they submit their transaction.
Even if you do away with orphans completely and have a deterministically ordered DAG with one transaction per block, the incentive is for users to reference as few parents as possible because every parent they reference will be given ordering priority over them, which is opposed to the timely confirmation incentive. The narrowest possible DAG is always desirable because the strength of the network decreases linearly in the number of partitions, and partitions are merged by referencing parents.
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
Btw, I commend monsterer and others on finally figuring out what I was stating upthread about DAGs and even monsterer's idea for a deterministic ordering.
|
|
|
|
sidhujag
Legendary
Offline
Activity: 2044
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 24, 2016, 11:42:43 PM |
|
Bringing this topic back on track, I'd like to discuss what I believe is the primary challenge in creating a decentralised, trustless, p2p cryptocurrency. Nash equilibrium for non professional miners.
Because power always centralises, if you reward miners, you create a point of centralisation so the apparent solution is not to reward them at all - instead you make it such that every user has to mine their own transaction in order to participate. And on the surface this seems fine but dig a little deeper and the nash-equilibrium is broken because the cheapest, least risky way (PoW cost and orphaning risk) for users to participate in this process is for them to ignore as much of the network as possible within the bounds of the protocol.
For example, (in a blockchain design with the LCR rule) if you require them to include as many mempool transactions as they can see in their blocks, they are incentivised to pretend they see only their own transaction because the probability of their new block being orphaned by double spend increases with every other transaction they include. This will cause the chain to become divergent because the structure of a chain doesn't permit high frequency blocks resulting from this rational behaviour; the LCR will be wildly selecting between candidate chains in an unintentional orphaning war caused by a combination of this rational behaviour, network latency and the chain structure itself.
In a probabilistic DAG design, even if you have only one transaction per block, orphaning risk still persists; submitting your transaction at the tip of the DAG is more risky than submitting it further down the hierarchy (because probability of being in an orphaned branch decreases with the number of children), so users have an incentive to widen the DAG which is the same as saying they agree not to make forward progress. In addition, users are incentivised to reference as few parents as possible, because the more parents they reference, the higher the chance of their transaction referencing a branch which will become orphaned due to a double spend which isn't apparent when they submit their transaction.
Even if you do away with orphans completely and have a deterministically ordered DAG with one transaction per block, the incentive is for users to reference as few parents as possible because every parent they reference will be given ordering priority over them, which is opposed to the timely confirmation incentive. The narrowest possible DAG is always desirable because the strength of the network decreases linearly in the number of partitions, and partitions are merged by referencing parents.
I maintain that unless there is a monetary incentive to behave in favour of the network as a whole, instead of being purely rational, the nash equilibrium is destroyed.
Thoughts?
Nash also said you win as a whole with the team and lose as an individual, therefor you must act in benefit of the entire network (team) to have the most efficient and effective system. This was his breakthrough where adam smith said the opposite. Nash equilibrium doesn't apply to society. It applies in some cases. Adam Smith did not say the opposite. In away he does.. nash argued a better global optima can ne achieved by the cooperation of a network rather than self-interest. Smiths model would achieve a local optima. Again Nash equilibrium only applies where the incentives are so aligned. It doesn't apply generally in society with unbounded entropy. Sorry please pay attention to what I write, so I don't have to repeat myself. A blockchain network of systems with unbounded entropy? doesn't sound very deterministic to me.
|
|
|
|
TPTB_need_war (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 24, 2016, 11:43:23 PM |
|
A blockchain with unbounded entropy? doesn't sound very deterministic to me.
You were talking about society. ![Roll Eyes](https://bitcointalk.org/Smileys/default/rolleyes.gif)
|
|
|
|
sidhujag
Legendary
Offline
Activity: 2044
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
February 24, 2016, 11:44:36 PM |
|
A blockchain with unbounded entropy? doesn't sound very deterministic to me.
You were talking about society. ![Roll Eyes](https://bitcointalk.org/Smileys/default/rolleyes.gif) Sorry please pay attention to what I write so I don't have to repeat myself ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
|