Why I support Jeff's BIP100

[-ck]

Lots of crosstalk then in this discussion, but I'd be very happy if bitcoin became so popular that we had 1GB of transactions regularly... Not sure where we'd find 1TB of transactions in a year let alone every 10 minutes, real or spammed.

[teukon]

Lots of crosstalk then in this discussion, but I'd be very happy if bitcoin became so popular that we had 1GB of transactions regularly... Not sure where we'd find 1TB of transactions in a year let alone every 10 minutes, real or spammed.

Yeah, I feel I could have been clearer.  Sorry if I mislead anyone.

I can't imagine demand for 1TBs worth of block space every 10 minutes but I believe that if we had the resources to provide this at very low prices that people will find a use for the space.

[jonny1000]

The way I see it:

• BIP100 measures demand
• BIP101 speculates supply

I believe both have problems and that we'd ideally like something which measures supply.

We are talking about how miners should set supply, we do not need to measure supply to do this.  What we need to measure is the price elasticity of demand and then set supply accordingly.  BIP100 encourages miners to do this, at first they may be poor at doing so, but over time miners will improve.

In the long run I don't think the upper bound can be safely removed without significantly changing the algorithm.

As demand increases, there will be pressure to increase the limit (to increase supply).  So long as miners can cope with the increased demand they have an incentive to vote to raise the limit for greater profits.  This is all well and good; we want these free-market efficiencies.

Miners may also have an incentive to vote for a lower limit, probably not in the first few years, but once the block reward is low they may vote to maximize revenue.  This would happen depending on what miners think about the price elasticity of demand.  Basically it depends on the demand curve, if reducing supply from S1 to S2 increases the price from P1 to P2, such that S1 * P1 < S2 * P2, miners should vote for a lower limit.

Illustrative demand curve with supply limit

Notes: Where the yellow and blue line cross is the equilibrium price, miner revenue is the area formed by a rectangle between this point and the point 0,0.  Miners can vote to lower the limit, if the demand curve is steep enough then revenue will increase when the limit falls.

[Nicolas Dorier]

BIP100 cut the competition between miners, so even if there are some miners that are able to spread blocks with higher size and faster, they will not be able to, because a cartel between miners, without risking so much, just a vote for a smaller size.

You can reverse the argument with BIP101.

"BIP101 cut the competition between miners, so if there are some miners that are not able to spread blocks with higher size and faster, they will be obliged to, because a cartel between miners, without risking so much, just push bigger blocks"

With BIP101 businesses will know that at a certain time they can be sure that there is a space to have blocks of certain size.
They will need for sure to know day by day which is the average size of the blocks, but they know that if there is the demand, some miners can still try to cover it.

Isn't it also the case with BIP100 ? if there is demand, miners will try to cover it thus increasing the size.

With BIP100, nothing is certain, every 3 month the size can go UP or DOWN (worst case)
So it is totally impossible to make any little plan for the future.

Even with BIP101 you can't be certain that miner will not adopt a small limit instead of a bigger one, so it change nothing.

The only valable argument I think of for BIP101 is about short term peak demand that BIP100 can't adapt for quickly. But this is an exceptional case.

[teukon]

The way I see it:

• BIP100 measures demand
• BIP101 speculates supply

I believe both have problems and that we'd ideally like something which measures supply.

We are talking about how miners should set supply, we do not need to measure supply to do this.  What we need to measure is the price elasticity of demand and then set supply accordingly.  BIP100 encourages miners to do this, at first they may be poor at doing so, but over time miners will improve.

What if the price elasticity of demand asks for a very high supply (i.e. Bitcoin takes off in a big way).  Will a miner vote for a block size limit that he himself does not have the resources to handle?

It seems like the BIP100 algorithm is trying to set a kind of "ideal supply" that would work well for the current demand.  What if we don't have enough resources to supply the "ideal supply"?

In the long run I don't think the upper bound can be safely removed without significantly changing the algorithm.

As demand increases, there will be pressure to increase the limit (to increase supply).  So long as miners can cope with the increased demand they have an incentive to vote to raise the limit for greater profits.  This is all well and good; we want these free-market efficiencies.

Miners may also have an incentive to vote for a lower limit, probably not in the first few years, but once the block reward is low they may vote to maximize revenue.  This would happen depending on what miners think about the price elasticity of demand.  Basically it depends on the demand curve, if reducing supply from S1 to S2 increases the price from P1 to P2, such that S1 * P1 < S2 * P2, miners should vote for a lower limit.

I agree (and this is a cool corollary of BIP100).  But notice that miner's are voting for lower block limits because demand is low.  What happens when these entrepreneurial miners correctly estimate that S * P is maximised at a point where S is very large?

[jonny1000]

The only valable argument I think of for BIP101 is about short term peak demand that BIP100 can't adapt for quickly. But this is an exceptional case.

BIP100 is likely to result in blocksize increases quite quickly at first, because when the block reward is high as there is little reason to vote for a lower limit, this should make it similar to BIP101 in the early years.

As time progresses and the block reward falls, if the network is fee driven it is likely to be able to handle short term demand variations much better, for various market driven reasons.  BIP100 can handle longer term variations.

For example if demand is high during peak shopping hours in the US, or at Christmas, the mining industry can allocate more power to mining and find blocks faster.  This is similar to how the national grid operates at the moment.  In the UK, when the national football team play an important match, at half time millions of national grid users make tea.  This causes a surge in power demand, however electricity producers predicted this in advance and respond by producing more energy.  Large amount of potential energy stored up in hydro facilities can be realized in one go, so the UK population can make their tea (http://news.bbc.co.uk/1/hi/uk/5059904.stm).  Bitcoin miners could also respond to demand surges and release large amounts of energy to increase their hash-power, temporarily producing faster blocks.

If we do not allow a fee market to exist, amazing features like this may not occur.

[HostFat]

You can reverse the argument with BIP101.

"BIP101 cut the competition between miners, so if there are some miners that are not able to spread blocks with higher size and faster, they will be obliged to, because a cartel between miners, without risking so much, just push bigger blocks"

No it's different, no miner is obliged to use bigger blocks, neither BIP100 or BIP101.
The miner can always chose to make smaller blocks than the actual version of the protocol can support.
It is NOT true the opposite.

The BIP100 can cut the possibility of some good miners to spread bigger blocks if the network can support them, because a minirity/cartel has chosen to.

Isn't it also the case with BIP100 ? if there is demand, miners will try to cover it thus increasing the size.

This is an assumption.

Even with BIP101 you can't be certain that miner will not adopt a small limit instead of a bigger one, so it change nothing.

This is true, but still there will be the space that someone can try to cover, with BIP100 it is NOT, the BIP100 can cut this possibility, because it opens an easy way to make cartels of miners.

The only valable argument I think of for BIP101 is about short term peak demand that BIP100 can't adapt for quickly. But this is an exceptional case.

BIP100 can increse the block size in 12 months to 32MB, by doubling it every 3 months, still it can cut the size, and this is the real problem.

[CounterEntropy]

Lots of crosstalk then in this discussion, but I'd be very happy if bitcoin became so popular that we had 1GB of transactions regularly... Not sure where we'd find 1TB of transactions in a year let alone every 10 minutes, real or spammed.

Have you checked BIP 1xx ?

[jonny1000]

What if the price elasticity of demand asks for a very high supply (i.e. Bitcoin takes off in a big way).  Will a miner vote for a block size limit that he himself does not have the resources to handle?

Yes, under BIP100 the blocksize limit could be very large if demand is large.  With respect to resources an upper bound could be in place for technical reasons.

Many BIP101 proponents make comments like "In 20 years your watch will be able to download the whole blockchain in seconds".  I have no idea if this is correct, if it is then maybe the technical limitations may become insifgnificant in relation to the market driven blocksize under BIP100.  This is why I think the stabilizers (on both sides) could be removed in the future.

[teukon]

The only valable argument I think of for BIP101 is about short term peak demand that BIP100 can't adapt for quickly. But this is an exceptional case.

BIP100 is likely to result in blocksize increases quite quickly at first, because when the block reward is high as there is little reason to vote for a lower limit, this should make it similar to BIP101 in the early years.

As time progresses and the block reward falls, if the network is fee driven it is likely to be able to handle short term demand variations much better, for various market driven reasons.  BIP100 can handle longer term variations.

For example if demand is high during peak shopping hours in the US, or at Christmas, the mining industry can allocate more power to mining and find blocks faster.  This is similar to how the national grid operates at the moment.  In the UK, when the national football team play an important match, at half time millions of national grid users make tea.  This causes a surge in power demand, however electricity producers predicted this in advance and respond by producing more energy.  Large amount of potential energy stored up in hydro facilities can be realized in one go, so the UK population can make their tea (http://news.bbc.co.uk/1/hi/uk/5059904.stm).  Bitcoin miners could also respond to demand surges and release large amounts of energy to increase their hash-power, temporarily producing faster blocks.

Exactly.  BIP100 can handle demand spikes as well as any other block size limit method (except arguably Meni's).  BIP101 has no principle advantage here.

If we do not allow a fee market to exist, amazing features like this may not occur.

While a fee-market focused limit might make reserve hashing power more commonplace, I fail to see why this in itself is a good thing.  Certainly, it is better to counter a surge with fast blocks than to not counter it at all.  Better still though would be to mitigate the surge by just having bigger blocks to begin with.  This way, people get their transactions confirmed even more quickly, and for a lower fee!

In a BIP101 universe, surges could be handled in the same way.

[Nicolas Dorier]

The miner can always chose to make smaller blocks than the actual version of the protocol can support.
It is NOT true the opposite.

The small miners can get kicked out of the market by a single or minority of miner which made a huge big blocks.
Also, BIP100 does not prevent a small miner to make a small block either, it only limit how a single miner can kick with big blocks.

Isn't it also the case with BIP100 ? if there is demand, miners will try to cover it thus increasing the size.

This is an assumption.

Isn't your's also ?

still there will be the space that someone can try to cover

If there is one or zero does not change a lot since only the average size count.

I don't get why you are so sure BIP100 allows cartel while 101 does not. It is actually the reverse with big blocks a big miners can kick the small one out of the market.

[teukon]

What if the price elasticity of demand asks for a very high supply (i.e. Bitcoin takes off in a big way).  Will a miner vote for a block size limit that he himself does not have the resources to handle?

Yes, under BIP100 the blocksize limit could be very large if demand is large.  With respect to resources an upper bound could be in place for technical reasons.

Yes, this is roughly what I'm arguing.  Growth is great but if we run into a problem with real resources then we'll need some kind of upper bound (or else mining centralisation).  So long as real supply is much larger than the "ideal supply" being calculated by BIP100 then things should work just fine.

Many BIP101 proponents make comments like "In 20 years your watch will be able to download the whole blockchain in seconds".  I have no idea if this is correct, if it is then maybe the technical limitations may become insifgnificant in relation to the market driven blocksize under BIP100.  This is why I think the stabilizers (on both sides) could be removed in the future.

Wow, sounds pretty wild.

My intuition is the other way around.  I expect humanity will demonstrate an apparently insatiable appetite for blockspace, just as it has for storage and bandwidth.  I myself could use 50 GB of blockspace were the price low enough.

I could be wrong.  It might be that block space becomes so vast and cheap that human desire simply can't keep up.  Block space would be as plentiful as air.  In this world, BIP100 would work fine but I'd still question the point of it.  Why, for example, would we want to artificially restrict the atmosphere to create an air market?  All I can think of is that without a fee market, transaction fees would tend to the largest miner's transaction-processing costs and the network security afforded by the fees would tend to zero.

[jonny1000]

My intuition is the other way around.  I expect humanity will demonstrate an apparently insatiable appetite for blockspace, just as it has for storage and bandwidth.  I myself could use 50 GB of blockspace were the price low enough.

This is a good point and I kind of agree with you, although supporters of BIP101 seem to disagree with this.  BIP101 supportrs typically think blocks may not be full.  This is why I advocate a lower and upper bound in BIP100 for now.  My point is that what if there is a stable optimal economic limit under BIP100, based on miner expectations of demand curves, that is much lower than the limit when bandwidth and other constraints begin to come into play?  This could happen, as although there could be an "insatiable appetite for blockspace", users may not be willing to bid a high enough fee.  Therefore miners vote for a lower limit than the elastic demand not supported by strong fee bids.

Lets consider a hypothetical scenario in 2035:
Size limit based on BIP100 voting: 2GB
Upper bound in the protocol: 8GB
Technical limitations, based on broadband speed and storage costs, ect, in 2035: 5TB

If we experience the above scenario for many years, we may decide we no longer need the upper and lower bounds, as we have high confidence in the voting process and high confidence the optimal miner voting strategy will not bring the limit near what can safely technically be done.

[teukon]

My intuition is the other way around.  I expect humanity will demonstrate an apparently insatiable appetite for blockspace, just as it has for storage and bandwidth.  I myself could use 50 GB of blockspace were the price low enough.

This is a good point and I kind of agree with you, although supporters of BIP101 seem to disagree with this.  BIP101 supportrs typically think blocks may not be full.  This is why I advocate a lower and upper bound in BIP100 for now.  My point is that what if there is a stable optimal economic limit under BIP100, based on miner expectations of demand curves, that is much lower than the limit when bandwidth and other constraints begin to come into play?  This could happen, as although there could be an "insatiable appetite for blockspace", users may not be willing to bid a high enough fee.  Therefore miners vote for a lower limit than the elastic demand not supported by strong fee bids.

Lets consider a hypothetical scenario in 2035:
Size limit based on BIP100 voting: 2GB
Upper bound in the protocol: 8GB
Technical limitations, based on broadband speed and storage costs, ect, in 2035: 5TB

If we experience the above scenario for many years, we may decide we no longer need the upper and lower bounds, as we have high confidence in the voting process and high confidence the optimal miner voting strategy will not bring the limit near what can safely technically be done.

Ok, let's see if I understand.

In this scenario, without the 2GB limit, blocks would not generally fill up.  This is not necessarily because no-one can think of any more uses for the space, but because the costs of relaying, processing, and storing any further transactions outweigh the benefits of using the space.  (I'll note that if this is true, we're probably operating under a different idea of "technical limitation").

With the 2GB limit in place, the total transaction volume is reduced but the total value of all transactions actually increases (following the S * P logic from earlier).

If this persisted for decades (as you suggest above 2 decades to 2035 + still more years of the same) I agree that it begins to look pretty safe to remove the 8GB limit.  In this case, I will have been practically proven wrong about my concerns.

Question: Why not remove both the 2GB limit and the 8GB limit?  With both limits gone, mankind will enjoy more transactions and lower fees.  Why should we prefer fewer transactions, higher fees, and a more complex protocol?

[jonny1000]

In this scenario, without the 2GB limit, blocks would not generally fill up.  This is not necessarily because no-one can think of any more uses for the space, but because the costs of relaying, processing, and storing any further transactions outweigh the benefits of using the space.  (I'll note that if this is true, we're probably operating under a different idea of "technical limitation").

This is not true no.  In my hypothetical scenario the costs of relaying, processing, and storing any further transactions are very low.  "Technically" the limit could be increased to 5TB and there would be no propagation or storage issues.  Because by 2035 my hypothetical smart watch has a 20TB/s connection and 500EB SSD or whatever.  The 2GB limit is imposed by a miner vote under BIP100 to maximize mining revenue.

Question: Why not remove both the 2GB limit and the 8GB limit?  With both limits gone, mankind will enjoy more transactions and lower fees.  Why should we prefer fewer transactions, higher fees, and a more complex protocol?

I think we may not need the 8GB upper bound limit in this scenario, but we still need the 2GB BIP100 voting limit.  The 2GB is imposed by miners to ensure fees are higher enough.  

Example Continued - The year is 2035

• 2GB BIP100 block limit
• All blocks are full
• $0.01 average fee

Miners then vote under BIP100 to increase the limit to 8GB, the situation changes to the following:

• 8GB BIP100 block limit
• All blocks are full
• $0.0001 average fee

Transaction volume is up, and users have more utility, but users are only prepared to pay $0.0001 per transaction, so the economic utility of these extra transactions for the users are low.  Mining revenue and therefore security has fallen by 25x (excluding the block reward).  This is why we prefer fewer transactions and lower fees, despite the fact "technically" the network could handle more.  Under BIP100 it is up to the miners to decide, in this scenario miners may think they have increased the limit too much and try to lower it again to increase average fees.  Remember more economic utility for users means the Bitcoin exchange rate may be higher, which would drive up mining revenue.  This would also be considered by miners when voting.

The above describes why BIP100 so brilliant.  Miners consider the economic utility of the transactions, to the user, by considering the fee the user will pay.  BIP100 may restrict the volume to a lower level than could technically be achieved, however this is done in a balanced way reflecting the impact on the bitcoin price and the utility users would get from these extra transactions.

[teukon]

In this scenario, without the 2GB limit, blocks would not generally fill up.  This is not necessarily because no-one can think of any more uses for the space, but because the costs of relaying, processing, and storing any further transactions outweigh the benefits of using the space.  (I'll note that if this is true, we're probably operating under a different idea of "technical limitation").

This is not true no.  In my hypothetical scenario the costs of relaying, processing, and storing any further transactions are very low.  "Technically" the limit could be increased to 5TB and there would be no propagation or storage issues.  Because by 2035 my smart watch has a 5TB/s connection or whatever.  The 2GB limit is imposed by a miner vote under BIP100 to maximize mining revenue.

Ok, my misunderstanding.

Question: Why not remove both the 2GB limit and the 8GB limit?  With both limits gone, mankind will enjoy more transactions and lower fees.  Why should we prefer fewer transactions, higher fees, and a more complex protocol?

I think we may not need the 8GB limit in this scenario, but we still need the 2GB limit.  The 2GB is imposed by miners to ensure fees are higher enough.  Mankind will not get more utility from more transactions than 2GB per second.  

Extend Example

• 2GB BIP100 block limit
• All blocks are full
• $0.01 average fee

Miners then vote under BIP100 to increase the limit to 8GB, the situation changes to the following:

• 8GB BIP100 block limit
• All blocks are full
• $0.0001 average fee

Transaction volume is up, and users have more utility, but users are only prepared to pay $0.0001 per transaction, so the economic utility of these extra transactions for the users are low.  Mining revenue and therefore security has fallen by 25x.  This is why we prefer fewer transactions and lower fees despite the fact "technically" the network could handle more.  Under BIP100 it is up to the miners to decide.  Remember more economic utility for users means the Bitcoin exchange rate may be higher, which would drive up mining revenue.  This would also be considered by miners when voting.

(emphasis mine)

Nice clear example.  I agree.  A fee market may cause profit-seeking miners to increase network security.

That said, I don't find network security to be a particularly compelling reason for having a fee market.  There seems to be practically no connection between a miner's incentives regarding voting and true demand for network security.  I could easily imagine this method yielding far too little security or far too much.  I believe there are better solutions to the large-block low-fee tragedy of the commons problem.

If each of the variables:

• The optimal level of network security;
• The point of unit elasticity (where S * P is maximal);
• The artificial limit derived by BIP100, driven by all of a miner's incentives;
• The limit of what a well-decentralised network can supply,

happen to have a particular relationship with one another then I'm sure BIP100 will behave superbly (even in the long term and without limits).  It wouldn't surprise me to learn that there is some subtle robustness (mechanisms naturally drawing these quantities into a healthy relationship) given the technical brilliance of the people that helped created the proposal but it's also possible that BIP100 is built upon a certain amount of wishful thinking.  I've warmed to it slightly thanks to your efforts but I remain cautious.

[Mashuri]

Transaction volume is up, and users have more utility, but users are only prepared to pay $0.0001 per transaction, so the economic utility of these extra transactions for the users are low.  Mining revenue and therefore security has fallen by 25x (excluding the block reward).  This is why we prefer fewer transactions and lower fees, despite the fact "technically" the network could handle more.  Under BIP100 it is up to the miners to decide, in this scenario miners may think they have increased the limit too much and try to lower it again to increase average fees.  Remember more economic utility for users means the Bitcoin exchange rate may be higher, which would drive up mining revenue.  This would also be considered by miners when voting.

The above describes why BIP100 so brilliant.  Miners consider the economic utility of the transactions, to the user, by considering the fee the user will pay.  BIP100 may restrict the volume to a lower level than could technically be achieved, however this is done in a balanced way reflecting the impact on the bitcoin price and the utility users would get from these extra transactions.

I don't think this takes into consideration competition.  Transactors will simply move more and more off chain, even to altcoins, if fees are kept relatively high by low block size limits.  Because of this, tx fees alone will not provide enough security for the network.

[teukon]

The above describes why BIP100 so brilliant.  Miners consider the economic utility of the transactions, to the user, by considering the fee the user will pay.  BIP100 may restrict the volume to a lower level than could technically be achieved, however this is done in a balanced way reflecting the impact on the bitcoin price and the utility users would get from these extra transactions.

I don't think this takes into consideration competition.  Transactors will simply move more and more off chain, even to altcoins, if fees are kept relatively high by low block size limits.  Because of this, tx fees alone will not provide enough security for the network.

Miners are driven to maximise the purchasing power of their income, not simply their nominal bitcoin income.  If most miners felt that a larger block size would yield greater real value then I don't see why they wouldn't vote for it.

Or, as jonny1000 noted:

Remember more economic utility for users means the Bitcoin exchange rate may be higher, which would drive up mining revenue.  This would also be considered by miners when voting.

[HostFat]

I still can't see any reason to give miners the vote to lower the size.
BIP100 can be "considerable" only if the vote to lower the size isn't present.

[jonny1000]

I still can't see any reason to give miners the vote to lower the size.

One simple reason is that without this option miners may be too relucant to increase the limit, as the decision could not be reversed.

Another potential reason is a cyclical fall in demand for using bitcoin.