Gold collapsing. Bitcoin UP.

[BldSwtTrs]

3. If I am wrong and the technological developments are not sufficient, then the marginal cost is likely to remain significant. Therefore we need an economically relevant block-size limit, in order to mitigate against the centralization pressure from these significant marginal costs.

[...]
2. IF the marginal cost is significant, we need a limit to mitigate against centralization threats



.

Could you please explicit the rationale behind what you are saying here regarding the need of a blocksize limit enforced at the protocol level to prevent centralization in case of significant marginal cost?

It looks like you are wrong.

 

[tynwald]

I think a better word than "zero" marginal cost of adding an extra transaction to a candidate block is "Insignificant" marginal cost. Let me try to describe my view again:

1. Historically, the marginal cost has not been insignificant.

2. In my view, in the future, the marginal cost is likely to be insignificant. The reasons for this are technological. For example, faster Internet connections, weak blocks, compact blocks, IBLT ect ect
There is a mining cost curve and miners lower down on the curve, may try to invest and driven up difficulty.

Assuming marginal cost per txn exists, the all miners will seek to optimize total revenue per block subject to natural limits + protocol limits + orphan risk. It's fallacy to contend that marginal cost is gt than zero and that its insignificant due to technical factors. These lessen but will never remove completely the cost vs benefit decision to be made on including each single txn.

In other words you can't believe that a "fee market" is necessary and good for Bitcoins survival while at the same time thinking that marginal cost doesn't matter. Micro-economics doesn't work that way, nor do the people who invest in mining as a business.
[doublepost=1468728543][/doublepost]

Could you please explicit the rationale behind what you are saying here regarding the need of a blocksize limit enforced at the protocol level to prevent centralization in case of significant marginal cost?

It looks like you are wrong.

Now you can see the internal contradictions of the Core business/technical model in plain view.

 

[jonny1000]

Could you please explicit the rationale behind what you are saying here regarding the need of a blocksize limit enforced at the protocol level to prevent centralization in case of significant marginal cost?

It looks like you are wrong.

For anyone reading this, please note the below is based on the assumption (which I do not necessarily agree with, that marginal costs of adding one extra transaction to a block are significant in relation to fee revenue, it is therefore not a valid counterargument to claim these costs are insignificant due to some technological improvement).

The main marginal cost is orphan risk cost. This is a concern as large miners do not have to propagate blocks to themselves. Therefore, the larger the miner the less the orphan risk cost relative to smaller miners. This causes centralization.

Please see the following example:

• A miner with 20% of the global hashrate only needs to propagate to 80% of the network. Therefore the orphan risk cost would be around 80% x cost.
• A miner with 0.1% of the global hashrate needs to propagate to 99.9% of the network. Therefore the orphan risk cost would be around 99.9% x cost.
• 99.9% x cost >> 80% x cost. Since we have already assumed the cost is significant in relation to fee revenue, this ensures we have a significant problem.*

Do you see how the larger miner has a comparative advantage? This would result in a scenario where miners compete over propagation technology and systems as opposed to hashing. Propagation systems themselves also have many economies of scale, which increase centralization. In contrast the economies of scale in hashing may fall over time.

There are also many other reasons (not necessary directly related to mining centralization) for wanting orphan risk cost to be low, such as:

• the "wasted work" problem
• ensuring a low orphan rate to improve the value of a confirmation and reduce double spend attacks

* Please note I assume the block reward is low, since it diminishes exponentially over time. I totally accept that if the block reward were to remain high, we would not have this issue.

 

[BldSwtTrs]

If this dynamic of hashing power/orphan risk turns out to be relevant in the real world (doubtful since there are lots of hidden complexities) then it exists regardless of the existence of a protocol enforced blocksize limit.

If we regard those two following assumptions as true:
- the orphan cost of smaller miner is higher than the orphan cost of bigger miner
- orphan cost increases with bigger blocks

Then to vindicate a protocol enforced blocksize limit, you need to explain why, in the absence of protocol limit, the orphan cost of smaller miners will increase more rapidly than their revenue (you probably cannot do that because it's probably a false assumption).
Because if, in the absence of a protocol enforced limit, the revenue of smaller miners increase more rapidly than their orphan cost, then their profit margin will increase, and this is the only things that matter regarding the decentralization question.

Problem with the "comparative advantage" frame is that it imagines mining as a closed market. The truth is capital can pour freely into the mining market and the only thing that matter to the capital is the profit margin. Therefore looking only at the cost part of the problem without looking at the revenue part is bad economics.

 

[BldSwtTrs]

In expanding market (with total revenue as the metric) profit margins increase alongside the number of competitors.

In declining market profit margins decrease and competition is driven by costs which is the ultimate centralization factor. Marx, among its numerous mistakes, at least understood that a declining rate of profit lead to big monopolies.

If you prevent smaller miners from increasing their revenue as fast as they can you are starting the countdown until their demise. If mining decentralization is the subject, instead of trying to set up a fee market it's better to care about increasing the total revenue of miners (because when revenue is expanding costs become less relevant).

 

[jonny1000]

If this dynamic of hashing power/orphan risk turns out to be relevant in the real world (doubtful since there are lots of hidden complexities) then it exists regardless of the existence of a protocol enforced blocksize limit.

If we regard those two following assumptions as true:
- the orphan cost of smaller miner is higher than the orphan cost of bigger miner
- orphan cost increases with bigger blocks

Then to vindicate a protocol enforced blocksize limit, you need to explain why, in the absence of protocol limit, the orphan cost of smaller miners will increase more rapidly than their revenue (you probably cannot do that because it's probably a false assumption).

The idea is that you set a blocksize limit small enough, such that orphan costs are so small they are totally irrelevant. With propagation technologies and speeds set to rapidly increase in the long term, perhaps in an exponential way, this should not be a problem.

• Remember when we used to think a 1.44MB floppy disk was pretty big??
  
• Remember when the 14Kb/s dial-up modem was considered very fast??
• ect ect...

In my view, it is absolutely critical, that by the time the block reward becomes low:

Orphan risk cost/fee revenue << ~0.1% (or some other small number)

The best way I can think of to achieve this is to have a blocksize limit. I am still an optimist, perhaps the limit could be over a GB is size and the equation could still hold true.

Perhaps many of you think I am just making this up, actually I came to this conclusion some time in 2011. I had also incorrectly assumed almost all others in the community had also reached the same conclusion, without speaking enough to people about this. I was clearly wrong in making this assumption, but I would guess others also made this mistake.

 

[freetrader]

If mining decentralization is the subject

It always becomes the subject when the discussion draws close to the core of the blocksize debate and its relationship to the business model of Blockstream.

If I were a Core developer or officer of Blockstream, I would hang my head in shame at the failure to keep Bitcoin decentralized, instead of spouting how the 1MB limit - a relic from Satoshi's days - is supposedly helping matters.

@Jihan talks about a failure of leadership - I must agree vehemently.

Currently Classic is at 3,6% miner support, and the miners who claim they as a majority supported 8MB blocks a year ago are not indicating even moderate support for 2MB (I'm not even talking about activation level support). There are some things that happened since that have not been fully explained.

 

[BldSwtTrs]

In my view, it is absolutely critical, that by the time the block reward becomes low:

Orphan risk cost/fee revenue << ~0.1% (or some other small number)

I am still an optimist, perhaps the limit could be over a GB is size and the equation could still hold true.

So why are you saying we need a blocksize limit to mitigate the centralization threat?

What we need is increasing mining revenue.

 

[jonny1000]

So why are you saying we need a blocksize limit to mitigate the centralization threat?

What we need is increasing mining revenue.

Whatever the revenue level, without a blocksize limit, in a competitive environment, miners will keep including more and more transactions into their blocks, up to the point where marginal orphan risk costs equals the fee of the marginal transaction. This maximises miner income.

The dynamics of this ensure that aggregate orphan risk cost is significant in relation to aggregate fee revenue. (In addition to this advanced wallets could estimate the approximate orphan risk cost of the transaction and reduce the fee to a slight premium over this level, why would any user want to pay more than that?)

 

[Inca]

When transaction fees make up the majority of their earnings, then miners will act in their best interests by attempting to maximise profit - i.e making blocks (as large as possible) containing as many transactions as possible.

The upper bound of the block size created by a greedy miner is limited theoretically by network orphan risk and importantly by transaction demand.

Thus, when (viable) miner income is comprised mainly of fees then: fees+profit = costs. If number of transactions is limited per block by orphan risk threshold and transaction demand then in a competitive mining environment costs will trend towards fees per block, and thus there is obviously a cost which can be calculated per transaction in each block. The use of 'marginal' is not helpful in this situation.

1. IF the marginal cost is insignificant, we need a limit to drive a fee market

But in the future when mining costs are largely covered by transaction fees then miners in free competition will set the lower bound of the transaction fee according to their own individual financial positions to secure their income. @jonny1000 you seem to be suggesting that miners are extremely stupid and are unable to include transactions into blocks which pay the most fees.

Quite a strange position, please explain this further as there does not seem to be any need whatsoever to engineer a fake fee market when a natural one exists already.

Finally, pretending that the 1mb blocksize limit was ever put in place for this purpose, or indeed to suggest that it is now providing this purpose is clearly batshit insane.

 

[jonny1000]

@jonny1000 you seem to be suggesting that miners are extremely stupid and are unable to include transactions into blocks which pay the most fees.

What!? When did I say anything like that?

The model is very simple, miners construct candidate blocks by first including the highest fee per byte transaction and then including more and more transactions, up to the point where the fee on the very last transaction equals the expected orphan risk cost of adding that bit of data to the block. This is a logical model, for maximising short term profit.

 

[Inca]

@jonny1000

The model is very simple, miners construct candidate blocks by first including the highest fee per byte transaction and then including more and more transactions, up to the point where the fee on the very last transaction equals the expected orphan risk cost of adding that bit of data to the block. This is a logical model, for maximising short term profit.

You are perfectly describing a natural fee market in action.

Now explain (ELI5!) why you need to create an artificial fee market above what you already described so well with the addition of a maximum blocksize limit.

EDIT: and keep it brief please, clarity and brevity go hand in hand if you understand what you are arguing..

 

[jonny1000]

Now explain (ELI5!) why you need to create an artificial fee market above what you already described so well with the addition of a maximum blocksize limit.

and keep it brief please

Ok I will try to give a simple (oversimplified) example. Lets assume the following:

• Block reward = 0
• Aggregate orphan risk cost / Aggregate fee revenue = 50%*
• Miner A has 50% of the global hashrate
• Miner B has 1% of the global hashrate

Therefore:

• The cost of orphan risk on miner A is 50% * 50% = 25% of total revenue
• The cost of orphan risk on miner B is 50% * 99% = 49.5% of total revenue
• All else being equal, miner A has a profit margin 49.5% - 25% = 24.5 percentage points higher than miner B
• This provides a large advantage to larger miners compared to smaller miners and we therefore have devastating mining centralization

With an economically relevant blocksize limit, ideally a non arbitrary, dynamic, market driven limit, we can avoid this problem by ensuring:

Aggregate orphan risk cost / Aggregate fee revenue << 0.5%

Is that sufficient clarity?

* Note: In reality this could be well over 50%, it is driven by two factors:
1. Miners will keep including more and more transactions in blocks, up to the point where the ratio is almost 100%, for the final most marginal transaction
2. Wallets will keep lowering the fee until the fee is only slightly above the expected marginal orphan risk cost of including the transaction

The above is just one of many reasons a high orphan risk relative to the size of the mining industry is bad.

 

[Inca]

@jonny1000:

You seem to have jumped from explaining why a blocksize limit is needed for keeping fees up (when a natural fee market exists) to now saying it is needed to prevent mining centralisation (too late).

Miners form into pools to prevent what you are describing from happening anyway, no?

EDIT: when i said 'needed for keeping fees up', I meant for the creation of an artificial fee market.

 

[jonny1000]

You seem to have jumped from explaining why a blocksize limit is needed for keeping fees up (when a natural fee market exists) to now saying it is needed to prevent mining centralisation (too late).

I never said I want to keep fees up, nor did I say I want a blocksize limit specifically for this purpose. As for metrics, I think its important to maximize the following:

• Total mining revenue = transaction volume * average fees * bitcoin price
• Aggregate fee revenue / aggregate orphan risk cost

It is true that higher fees may help maximize these things, however these things can also be maximized by the other variables in the respective equations. I do not look at the situation for a single point of view and do not only focus on fees. I do think fees can fall from current levels as technology improves, however we still need an economically relevant blocksize limit to ensure that the metrics above do not fall too low if fees fall to near zero.

Remember in a competitive market price = marginal cost, if technology improvements ensure orphan risk cost (marginal costs) falls to very low levels (almost zero), then fees will also fall to almost zero. We need to prevent this to ensure the network remains secure and robust.

Miners form into pools to prevent what you are describing from happening anyway, no?

No, I do not see how pools prevent this problem.

 

[awemany]

[...]
2. In my view, in the future, the marginal cost is likely to be insignificant. The reasons for this are technological. For example, faster Internet connections, weak blocks, compact blocks, IBLT ect ect
[...]

One side sees it as insignificant and dangerously low for 'full node decentralization', and the other side says it is cheap - meaning Bitcoin can scale well to many transactions ...

At this point in time, talk about making sure to 'keep up miner fees' is IMO at best severely misguided, and at worst destructive propaganda: We are far from a situation where we have grown to maximum permissible on-chain transaction rate. Even if you are worried about FNC.

Bitcoin will have failed if it is only the 2nd-largest coin by market cap - because it won't stay there for long. This is the kind of danger we are facing now.

And it is further my impression is that as soon as we are at a level where we are politically significant, bigger blocks and node size / node count are not as much an issue anymore, because communities/countries start to want to have their own full node, too. An incentive effect which is currently only playing out at the level of dispersed individuals ...

(And individuals are much easier to get in line than lets say a whole city that decides to run a (bigger, bigger blocks) full node ... )

[...]
I am certainly not guessing demand for blockspace will increase due to payment channels. For the record I am skeptical of side chains but I think the LN is a very intelligent idea, which may become successful in many years time. I certainly would not assume LN will be successful, I agree that would be irresponsible. My point was as follows:

1. If LN fails then its insignificant and is likely to have no impact on demand
2. If LN succeeds its impact on demand could be approximately proportional to its success.
[...]

LN as an additional feature might increase the total Bitcoin transaction demand - and thus also market demand of Bitcoin. I am not disagreeing with that (I don't really know what went down regarding this discussion here in the last couple days ... I kind of missed that). But at the same time, LN can in principle also cannibalize on-chain transactions. I think most here are not opposed to LN, but are opposed to forced cannibalization of on-chain transactions to LN hubs.
There is a fine line - and maybe even a trade-off - between LN channels being a nice, organic addition to Bitcoin (and they might even turn out to capture most of the additional transactions, should they be successful) - and trying to force users into LN, because dammit, I am feeling particularly dictatorial today.

If I look at the current Core/BS landscape, I see this:

They have talked about and have used their various means of power to apply the stick to Bitcoin users, to 'educate them that Bitcoin transactions are not free' and so forth.

There is the extremely damaging hubris and arrogance of trying to apply the stick in an open, voluntary world, where one is free to walk away (to Alts) at any point.

Watching this as someone who likes a healthy, innovative Bitcoin ecosystem hurts quite a lot.

But apart from that, what they have not at all applied, yet, is the carrot.

Only Blockchain's recent release of their LN infrastructure has features of that. Without being very deep in the technical details of all this, I imagine that e.g. network protocols for LN transaction forwarding etc. could be documented, written down. Wallets could be encouraged to implement a common standard.

I have seen few attempts from Blockstream to play on that level, so far.

They continue to do their best to drive people away from LN, potentially causing the whole community to throw the baby out with the bathwater if Jihan and the other miners decide to reject core.

If you are a LN supporter, I fail to see how you can view that in a good light.
And regarding SegWit: Worries about bandwidth, storage and CPU time for validating transactions do not lead a good engineer to this beast.

For example, the built-in space discount is something a sane Core dev team would have avoided and put back to the miners to decide. They will decide on the exact trade-off anyways, so what good is it to have such a weighting in the software, other than unnecessary complexity and a (further) try at elevating oneself to be the authority in all things Bitcoin?
[doublepost=1468761679][/doublepost]

The main marginal cost is orphan risk cost. This is a concern as large miners do not have to propagate blocks to themselves. Therefore, the larger the miner the less the orphan risk cost relative to smaller miners. This causes centralization.

And on this, before I forget: Who sells and who buys in a market is a matter of perspective!


The miners have a demand for transactions!


The transactions will arrive worldwide, and not from a single point. I am quite certain, though cannot prove yet, that a 'world miner' is not the natural equilibrium here.

 

[cypherdoc]

kinda funny watching the shitfest going on over on reddit regarding censorship. this thread has never had anyone banned nor any post removed. i do recall one or two blatantly obvious advertising posts (like gambling or new altcoin or buy this, buy that) that Bloomie removed that were clearly spam. but that's it!

 

[yrral86]

Remember in a competitive market price = marginal cost, if technology improvements ensure orphan risk cost (marginal costs) falls to very low levels (almost zero), then fees will also fall to almost zero. We need to prevent this to ensure the network remains secure and robust.

No, market price is marginal cost + amortized fixed cost. Unless miners have free hardware and free electricity, fixed costs will never be zero.

 

[xhiggy]

I guess your comment here is technically true, a ponzi scheme is where an operator pays returns to investors from new capital by new investors, rather than from profit earned through genuine operations. I guess Bitcoin does not fall into this category, because there is no operator and no returns are paid.

However, I think Stolfi is fair to contrast Bitcoin with a ponzi scheme as Bitcoin does appear to share some characteristics with them, depending on how you look at Bitcoin. For example you could argue that if nobody used Bitcoin, then there would be no miner fees and no demand for bitcoin to pay the miner fees, then miner incentives would fall to low levels and the system would collapse. Therefore one could argue Bitcoin falls to zero if people do not use it, which is kind of like a ponzi scheme, but I guess you could argue it is the same for almost all businesses.

Another point of view is that Bitcoin is unique in that its immutable and robust ledger has intrinsic and inherent value, providing utility to customers. There is therefore genuine demand for bitcoin to pay miner fees, which enable one to use the ledger. These miner fees provide an inherent usage for bitcoin, that one could argue distinguish it from a ponzi scheme.

I hope this explains to you why a fee market and fees are so important to many in the community. It is one of the key characteristics which provides inherent value to Bitcoin. These features are not necessary normal or direct characteristics of money (store of value, unit of account, medium of exchange). Fees therefore go right to the core of establishing bitcoin as money. For example many argue gold is directly useful as money (store of value, unit of account, medium of exchange), but gold also has inherent value (e.g. jewelry, electronic conductor, teeth ect). In my view the definition of inherent value in this context is non monetary uses. I think these non money uses for gold play and have played a key role in helping establish it as money in the past. I think the same may be true for transaction fees in Bitcoin.

Unfortunately there seems to be a lot of division in the community about the philosophy of fees in the system. I really hope we can debate these issues in the future in a more respectful way.
[doublepost=1468560731][/doublepost]For example:



Please do not call the limit "retarded". The limit is a crucial and integral part of the system to many in the community. How do you seek to persuade them if you do not respect them and imply they may be retarded? Why not say that you are trying harder and harder to understand the many valid reasons for the limit, but on balance it might be a good idea to increase the limit. One could then have a patient and polite debate about the issue, while respecting the rights and ability of others to veto any increase if these people are significant in number and they cannot be persuaded.

Fees are a part of bitcoin of course, their magnitude doesn't need to be dominated by the blocksize for bitcoin to function.

 

[AdrianX]

Does anybody actually argue that the marginal cost of including an additional TX is zero? Isn't it blindingly obvious that that can't be the case? I'm not a physicist but wouldn't that violate one of the laws of thermodynamics or something?

What I've come to understand is the cost of including a transaction with a technology like Xthin continuously approaches zero as the risk of block orphans decrease as the likelihood the transaction relayed to all nodes and miners increase.

A technology like RBF and Core's (0.12.0) Memory pool limiting conversely increase the risk.