Topic: [If tx limit is removed] Disturbingly low future difficulty equilibrium - page 7. (Read 37667 times)

This bears repeating again, apparently.


The increase from 50 to 150 BTC per block is to accomodate the future tripling in money supply. The incredibly large fees are necessary to maintain a constant security level (e.g. a constant ratio of attack cost to market capitalization). Yes, you could use a mixture of crippling txn fees and extortionate demurrage fees (or equivalently perpetual money creation). Yes, I agree that the inflation with txn fees product is less toxic than the pure inflation or pure txn fee products. Nevertheless, a mixed product is already available here: www.paypal.com. Won't bitcoin will be late to market?

More seriously, shouldn't we be trying to create a proof-of-stake system which potentially solves the security problem without requiring truckloads of money to be distributed every 10 mins.
To get at the problem you need to leverage the money already invested in bitcoin when you motivate the txn verifiers to tell the truth. Under the current txn fee / inflation systems txn verifiers are bribed with money to stay honest. Alternatively, a well-designed proof-of-stake system could threaten the btcs holdings of people if they lie. If you can confiscate money from people who behave dishonestly, you will not need to supply a mandatory participation system with any outside payments. You will already have txn auditors by the balls, so why pay them. If participation is voluntary, you will almost certainly need some outside payments, but I'm pretty sure that you could cut these down on these dramatically. There are probably many ways of organizing this. We should be trying to figure out a system which creates strong incentives to tell the truth (that is to not attack bitcoin) with a minimum investment of outside resources (eg. txn fees / inflation levies). Everyone needs to be a little open-minded during this process because there are probably many ways of doing this (some much better than others). The first step is to agree that there is a problem (like AA). The second step is to admit that txn fees or inflation aka demurrage can't fix this problem.

Why do miners need 150 btc reward with each block? It depends on the value of BTC, right?
Of course, a 37% annual demurrage rate would be unacceptable. Here you have some calculations assuming 3% demurrage rate for freicoin.
With any demurrage rate and any block reward, the monetary base will eventually converge to a fixed amount. Also lost wallets will automatically be recovered by the network.
With demurrage you can also have tx fees, and there's going to be more transactions with demurrage, as users are discouraged to store the money for long periods of time.

This bears repeating. It seems to escape much of the readership.

The future (and very near term or we are bust) I consider is when tx fees are dynamically defined not by a human, but by the system itself depending on variety of factors evaluating system status. Miners will dynamically get information about that and decide if they want to participate in mining the next bitcoin block or not depending on their mining efficiency.

Some illustrative calculations with using different assumptions:

Assumption 1: The 1 MB txn limit is binding for some reason.

Security should be measured by the ratio of the reward per block to the money supply. Currently, "security" is about 50/7,000,000= 7.1 * 10^-6
If there are 21 million instead of 7 million BTC in the money supply, the block reward will need to triple to 150 in order to maintain security, i.e. 150/21,000,000 = 7.1 * 10^-6
1 MB maximum is equivalent to about 4000 transactions per block. In order for 4000 transactions to yield 150 BTC, you would need a per txn fee of 150/4000=0.0375. Currently, 0.0375 BTC is worth about USD$0.50. This seems alright, but it could not last. With growth in the demand for txns, BTC will appreciate, leading to a much larger txn fee. For example, right now I doubt that there are even 500 txn per block.
If the # of txns per block increases by eight fold or more to reach 4000, we should expect at least an eight-fold appreciation of BTC vs. USD.  This would increase the txn fee to USD$4.00.
Such a high txn fee would greatly undermine the usefulness of the technology.

Assumption 2: Assume the txn limit can be increased so as to maximize txn fee revenue. VISA processes about 1 million txn per 10 minutes (i.e. per block).  Suppose Bitcoin makes each block 250 MB and handles the same txn velocity as VISA. Then a 150 BTC per block reward can be maintained with a txn fee of 150 BTC per block / 1 million txn per block = 0.00015 BTC/txn. This doesn't sound sound like much until you think about how much BTC will be worth under this scenario. Particularly important is the ratio of BTC's market cap to its txn volume. The more money is hoarded, the more onerous the txn fees need to be to protect security. I consider two possible BTC valuation scenarios . They are designed to represent upper and lower bounds on BTC market cap.

Worst Case Scenario: BTC continues to be used as a store of value rather than as a txn medium. This makes it hard to raise money from a txn tax. In this scenario, I assume that BTC keeps its current ratio of market cap to txn velocity. Currently, market Cap is around 100,000,000 USD. Currently, txn Velocity can be approximated as double the trade volume on Mt. Gox, or about 500 BTC per 10 minutes. Suppose that these 500 BTC of txns are made of 100 txns with an average size of 5 BTC occurring every 10 minutes. We then have 1 million USD market Cap per txn per 10 minutes. If BTC grows to handle 1 million txns per 10 minutes instead of 100 txns, then market cap would increase ten thousand-fold to 1 trillion USD. This would yield a value per BTC of 1 trillion USD / 21 million BTC = 50,000 USD/BTC. Remember that the tax per txn is 0.00015 BTC. This tax is equivalent to 7.5 USD per txn.

Best Case Scenario: BTC becomes a txn processor like Dwolla or VISA rather than a store of value. Like VISA, is BTC purchased on as needed basis to make txns. In this case, maybe we should assume that the BTC market cap grows to only VISA's market cap (~63 billion) rather than to 1 trillion dollars. This yields a value per BTC of 63 billion USD / 21 million BTC = 3000 USD/BTC. Given the tax per txn of 0.00015 BTC, this is equivalent to USD$0.45 per transaction. Slightly higher than Dwolla's rate of USD$0.25 per transaction.


In the worst case scenario, bitcoin appears doomed. Even in the Best Case Scenario, it is not clear that bitcoin will be competitive with other txn processors. Some people have proposed taxation of hoarded BTC to discourage excessive reliance on bitcoin as a store of value. Such a tax would ensure a 'best case scenario' where BTC is used like VISA/Dwolla rather than a store of value. Unfortunately, the tax on wealth holdings necessary to maintain security is exceptionally large. To generate a 150 BTC per block reward for miners, you would need to tax stored wealth at a rate of 37% per annum. Such a high tax rate is unlikely to be accepted.

This dire situation could probably be turned around using a proof-of-stake system or a hybrid proof-of-stake and proof-of-work system. The advantage of proof-of-stake or hybrid proof-of-stake/proof-of-work is that you can attain an equivalent security level using a much smaller reward.

if bitcoin were the currency in a trillion dollar economy, then the people for a few hours a day should help secure the network they use. at most it would cost each person a few dollars a month, less than what you pay out your ass in fees that visa and friends charge. as a matter of fact, the sheer volume of cpus out there alone far exceeds any means that a company would be able to buy enough gpus. if only 20million cpus were mining at 1mh/s thats close to 20th/s. if on average people only mined 4 hours a day.

according to steam, the most popular gpu is the 9800, most of them have a hash rate at or above 30, the 9800 has ~5% of all steam users.
the 8800, most have over 20, and many over 30
the gtx260 ranges from 40-70
the 5770, at ~4% gets ~200+

lets do some averaging, shall we.

ill throw in some low end gpus to even it out and do less work.

gts150 lower than 5
8800 25
260 50
5770 210
5850 300
5870 450

thats around 173mh/s per person, assuming they do gpu mining so even if only a million people were to mine, that's around 150 th/s

yes the numbers are half assed, but it does give a small taste of the untapped potential of the average consumer computer. the census says that 44million people have computers with internet access. this is with only 1 million people with less than modest machines that would cost 700-1200 usd. thats only 2% people. only 2% of the US population would need to mine to insure that bitcoin is secure against medium sized threats. but yeah it would be nice to have that hash rate number to see how much power you would need to get 50%+

This is true.
Maybe when it starts (I read is developed and will start soon) we can make more accurate predictions about how it affects security.
I think this is going to benefit mainly namecoin, but also bitcoin, since the reward for miners will be higher.
I guess it will also affect the price of both currencies.

With merged mining, the total hashing power is not determined by incentives of Bitcoin blockchain alone, but combined incentives of all possible blockchains. Not saying that it's the final solution to any problem, but just something to keep in mind when talking about future incentives and hashing power.

I am assuming that in the future the only way to significantly increase the value of Bitcoin will be its use for more trade, more trade means more transactions, more transactions means more secure network. Yes hording can bring the price up as well, but by the time Bitcoin reaches a large user base and economy it will not be prone to rapid value increases as it is currently, leaving speculators to move onto other higher gain investments, as such to significantly increase the value without speculation you would have to expand the user base, an expanding user base would be a very slow increase in value, in fact it might scale protection faster than scale the exchange rate.

If the fee scales down with increasing exchange rate it means that the level of protection does not scale with the total value of the bitcoins it is protecting. That makes the potential gains from attacks higher without increasing the cost.

A $1 trillion economy does a lot more than just 2000 transactions per second. And even then, my proposal is based on everyone paying a fee of about 1/10th of a cent (exchange rate does factor in, but then the BTC numbers go down and the value stays the same) if you wanted 10 times more security with only 2000 tps you could bring the fees up to a measly 1 cent. That would pay for 240 times the security of the current network, at only 2000tps. The larger the economy, the more transactions, the more money paid for protection. It scales. 

The current network hashing capacity is enough for securing a $100M economy. It is not enough to secure a $1T economy. It can be overrun by anyone with a multi-million dollar incentive to do so, be it double-spending of huge transfers, political sabotage, shorting, whatever. As Bitcoin grows its security requirement will grow, and I'm not sure it will be achievable with trivial transaction fees alone.

Thats why i measured it all in BTC, trying to guess the exchange rate is pointless, you can change all the numbers here, they scale with the exchange rate if it goes up the fee goes down, same protection, my point was that once we reach that sort of volume transaction fees can easily pay for network security at a very very low cost per transaction.

that can be solved by only paying out when the person has reached a certain amount. they do this now, but if more considerations were taken they would take up even less space.

I think the current limit is 1MB (this can easily be raised). The smallest transactions are 258 bytes; while transactions can run over 1KB, especially things like pool payouts. So you're currently looking at an upward limit of around 4,000 transactions in a block.

I'll eidit this to more thorougly correct your mistakes later, but for now point out the most important one. You are assuming that the bitcoin to USD exchange rate will remain fixed even if bitcoin grows to handle as many transactions as VISA.

To help me come up with estimates... What is the limit on the number of transactions bitcoin can process per block (the limit on max block size converted into a count of txns.)

The idea behind transaction fees. is that by the time 2040 rolls around Bitcoin will have already failed, or be successful enough that the transaction volume can support the network with minimal fees. Currently 201600 BTC are issued per month for Bitcoin protection. And no, the current transaction rate could not support that same amount of security, if we assume that by 2040 Bitcoin will have either already failed, or be at least have as much trade as Visa handles currently, about 2000 transactions per second. At 2000tps if each trade pays a fee of .001 BTC (about 1/10th of a cent)  that adds up to 2 BTC per second, 120 BTC per minute, so at 10 minutes per block that's 1200 BTC per block reward. Thats 24 times more "network security" than we have today, at a trivial fee per transaction.

I see, it was just a misunderstanding.
By the way, you're the second person that tells me he likes the idea of freicoin. I think it doesn't make much sense to start it right now.

They don't need to identify anything. They just need to sign a block hash once a day. It's the duty of receivers of large transactions to wait until the block receives a proof-of-stake before considering the money safe.

Because the cost of proof-of-stake is close to 0, it doesn't matter much if the incentive is weak. And it should be straightforward to have mechanisms to make it easier and to deal with the contingency of a large number of coins not voting anyway.

yes, in the current system if too many people hoard, then the coins are almost automatically considered dead by the community if done for years at a time. then things would cost less btc because there is less of it. im sure a simple algorithm could tell how many coins are "active" based on how many blocks ago they were traded.