Topic: Changing the re-targeting rules with a soft fork (Read 826 times)
I proposed something related but got moved to alt-coin section: https://bitcointalksearch.org/topic/m.2739332
At first thoughts, this would be impossible. However, the rule for a soft fork is that a majority of the hashing power must accept the more restrictive rules. Increasing the difficulty is a more restrictive rule, so it could be a hard fork.
In this case, the difficulty itself is what is being modified. This makes it slightly more complex than the 51% rule.
Doubling difficulty
If the new rule was that difficulty was doubled, then the threshold for acceptance is 2/3 of the mining power.
If 75% of the miners agreed with the doubled difficulty, then the high difficulty fork will have an effective hash rate of 37.5% (since by definition, they have to throw away 50% of their blocks).
The standard difficulty fork would have a hash rate of 25%, so the high difficulty fork would increase faster, even though it is higher difficulty.
Miners who are producing blocks for the high difficulty fork wouldn't be able to use them on the low difficulty fork, since the previous hash pointers wouldn't match.
When retargets happen, they will be based on the higher difficulty, so it would still be one block per 10 mins. This means that it doesn't actually double difficulty (except for 1 retarget period).
Variable difficulty
Many alt coins retarget after each block. This protects the network against sudden loss of mining power.
An alt coin effectively pays (mint fee) / (difficulty) per hash to miners. This means that modifying the difficulty changes how much the network is paying for hashes.
For the main network, this is not that big a deal, since the bitcoin network if the primary source of demand. If difficulty goes up, then market price falls, since the main network "buys" most the hashing anyway.
Alt coins, on the other hand, operate in a competitive hash environment. The difficulty of an alt coin has no influence on the market price for hashing power.
If an alt chain is paying 120% of market price for hashing, it will get lots of miners. If the retarget causes the difficulty to jump up by 4, then suddenly, it is paying 30% of market price for hashing. This causes many miners to leave.
If the difficulty is adjusted by 1% per block, then the difficulty will drop more slowly, 120%, 119% ... 101%, 100%, 99%. Some miners will leave when it hits 99%, but there is less likelihood of a mass exodus. When they leave, the difficulty will drop for the next block and the network will pay more coin per hash. This keeps the alt coin's payments per hash at around market price.
Implementing this for bitcoin could be implemented as a soft fork. There wouldn't be full flexibility. The difficulty would have to be restricted to greater than the standard difficulty. However, subject to that, it could float up and down.
If more than 80% of the miners agreed the range could be 1.0X to 4.0X. This might be a better target, since 4X is the maximum change in the difficulty.
This makes it possible to separate old clients from the network with only 20% of the hashing power, so it introduces a different vulnerability. However, once implemented, newer clients would be coded to accept the new hashing difficulty rules after block N.
There could also be an "exit" clause, which checks if an alternative chain is growing faster and drops difficulty. This may lead to other vulnerabilities.
This change would protect the bitcoin network against "hyper-transfer".
In a competitive world, you can't drop you price in 4X steps and hope for the best. 99% of market price might get you 90% as much hashing as 101%, but 30% will get you almost nothing.
In this case, the difficulty itself is what is being modified. This makes it slightly more complex than the 51% rule.
Doubling difficulty
If the new rule was that difficulty was doubled, then the threshold for acceptance is 2/3 of the mining power.
If 75% of the miners agreed with the doubled difficulty, then the high difficulty fork will have an effective hash rate of 37.5% (since by definition, they have to throw away 50% of their blocks).
The standard difficulty fork would have a hash rate of 25%, so the high difficulty fork would increase faster, even though it is higher difficulty.
Miners who are producing blocks for the high difficulty fork wouldn't be able to use them on the low difficulty fork, since the previous hash pointers wouldn't match.
When retargets happen, they will be based on the higher difficulty, so it would still be one block per 10 mins. This means that it doesn't actually double difficulty (except for 1 retarget period).
Variable difficulty
Many alt coins retarget after each block. This protects the network against sudden loss of mining power.
An alt coin effectively pays (mint fee) / (difficulty) per hash to miners. This means that modifying the difficulty changes how much the network is paying for hashes.
For the main network, this is not that big a deal, since the bitcoin network if the primary source of demand. If difficulty goes up, then market price falls, since the main network "buys" most the hashing anyway.
Alt coins, on the other hand, operate in a competitive hash environment. The difficulty of an alt coin has no influence on the market price for hashing power.
If an alt chain is paying 120% of market price for hashing, it will get lots of miners. If the retarget causes the difficulty to jump up by 4, then suddenly, it is paying 30% of market price for hashing. This causes many miners to leave.
If the difficulty is adjusted by 1% per block, then the difficulty will drop more slowly, 120%, 119% ... 101%, 100%, 99%. Some miners will leave when it hits 99%, but there is less likelihood of a mass exodus. When they leave, the difficulty will drop for the next block and the network will pay more coin per hash. This keeps the alt coin's payments per hash at around market price.
Implementing this for bitcoin could be implemented as a soft fork. There wouldn't be full flexibility. The difficulty would have to be restricted to greater than the standard difficulty. However, subject to that, it could float up and down.
If more than 80% of the miners agreed the range could be 1.0X to 4.0X. This might be a better target, since 4X is the maximum change in the difficulty.
This makes it possible to separate old clients from the network with only 20% of the hashing power, so it introduces a different vulnerability. However, once implemented, newer clients would be coded to accept the new hashing difficulty rules after block N.
There could also be an "exit" clause, which checks if an alternative chain is growing faster and drops difficulty. This may lead to other vulnerabilities.
This change would protect the bitcoin network against "hyper-transfer".
In a competitive world, you can't drop you price in 4X steps and hope for the best. 99% of market price might get you 90% as much hashing as 101%, but 30% will get you almost nothing.
Jump to: