Block space as a commodity (a transaction fee market exists without a block size limit)

[Peter R]

The purpose of this thread is to discuss and debate the idea of block space as a normal economic commodity and topics related to the Bitcoin transaction fee market.

Background material:

1. Fee Market paper
2. Fee Market talk

Abstract (from not-yet-published first revision to paper)

This paper shows that a Bitcoin transaction fee market would exist without a block size limit provided the inflation rate is nonzero and more than one miner or mining pool exists. To show this, the paper first introduces the block space supply curve and the mempool demand curve. The former describes the cost for a miner to supply block space by accounting for orphaning risk. The latter represents the fees offered by the transactions in mempool, and is expressed versus the minimum block size required to claim a given portion of the fees. The paper explains how the supply and demand curves from classical economics are related to the derivatives of these two curves, and proves that producing the quantity of block space indicated by their intersection point maximizes the miner’s profit. The paper uses this result to show that an unhealthy fee market—where miners are incentivized to produce arbitrarily large blocks—cannot exist since it requires communicating information at an arbitrarily fast rate. The paper concludes by considering the conditions under which a rational miner would produce big, small or empty blocks, and by estimating the cost of a spam attack.

 

[Justus Ranvier]

Proving that the supply of any service which requires the expenditure of time or energy to produce is always finite is a bit like proving that a perpetual motion machine isn't.

Anybody who thinks they've invented a perpetual motion machine is invariably overlooking some source of energy which keeps the machine operating, and likewise anyone who thinks they've found a service that requires scarce resources to produce yet that has no supply curve is invariably overlooking a cost somewhere.

 

[Peter R]

Quite true. One could probably analyze the fee market from another perspective (resources required to store the UTXO set, for example) and arrive at another supply curve. The effective supply curve is of course the net effect of all supply curves. I'm confident that presently orphaning risk is the dominant production cost.

 

[Justus Ranvier]

There should be a supply curve based on the cost to the miner for propagating their block across the network.

Right now that cost is being subsidized by uncompensated full nodes due to incomplete/poor P2P network design.

This free rider problem is one of the few arguments against increasing the block size which I agree with, but it's rather strange that when I propose solutions to the problem so that it's no longer a barrier to increasing or removing the production quota, suddenly all the people who bring it up don't appear to have the slightest interest in solving it.

 

[Mashuri]

I watched your S.B. presentation and agree that's it's the correct path for block size. One thing to account for is the current bitcoin relay network all the mining pools and big nodes currently use. Greg Maxwell likes to reference that as a counter to block propagation / size limitations. I think a market can still exist since, if a miner creates a monstrously huge block, other miners will be incentivized to create a competing smaller block instead mining of on top of it, if they believe they have a good enough chance to orphan it on the standard, P2P network.

 

[Peter R]

Thanks for joining the discussion, Mashuri!

I think schemes like the Relay Network to implement coding gain (the "gamma" variable in my model) are what is necessary to reduce the network propagation impedance to make it more affordable for miners to produce larger blocks in the future.

For example, right now it would be very costly to produce a 128 MB block, for instance:

 

[Mashuri]

Where did you get that chart? Is it in a report of yours?

 

[Peter R]

@Mashuri : yes, it's on page 12 of the fee market paper (refer to link in the first post of this thread).

 

[cypherdoc]

@Peter R

have you heard exactly what type of "pseudo-compression" scheme Corallo utilizes in the relay network?

 

[Georg Engelmann]

Great talk Peter R. I'm going to read the paper later.

I think that ideally a block size limit shouldn't exist. BIP101 is a compromise for those who are afraid of having no limitation.

 

[Peter R]

@Peter R

have you heard exactly what type of "pseudo-compression" scheme Corallo utilizes in the relay network?

My understanding is that they relay the TXs included in the solved block using a 2-byte index (they sort of send a "map" of how the TXs are sorted in the block). This only works for the TXs the other miners know about already. I believe there is some procedure for relaying the TXs in full that the receiver may not have in his mempool. I think the theoretical coding gain for average TXs is about 250x (500 bytes -> 2 bytes) if the mempools are homogenous, and significantly less as mempool heterogeneity increases.

 

[molecular]

Thanks for making this thread.

The paper says that blockspace cost is proportional to Bitcoin's inflation rate (R/T). I'm not sure I can intuitively understand this and also I cannot spot exactly where this is shown (it's mentioned in the conclusions). I didn't read all parts of the paper yet, so maybe I just missed it.

I'm also unsure about the implications: what happens when inflation rate tends to zero with the block reward?

 

[molecular]

@Melbustus brought up the issue from my previous post in the gold collapsing thread:

http://bitco.in/forum/threads/gold-collapsing-bitcoin-up.16/page-31#post-950

Quite an interesting discussion followed

 

[Yoghurt114]

Hi,

I enjoyed your paper and your talk (ignoring the needless sneer at the end, but ok)

My primary concern with this paper is that the economics implied, I feel, would make for blocks that are easily much, much larger than many auditors of the blockchain, nodes - in the form they exist today, can hope to keep up with. Even if they could, their ability to continue to keep up would be wholly outside their control; if miners, for any reason, reduce orphan risk in any way whatsoever (IBLT, relay network, or similar (near-)constant-time block propagation), block sizes would naturally increase (assuming a fee market has indeed formed, of course) and auditors would presumably have no way to anticipate or react to that - possibly leading them to stop their auditing.

To me, it is entirely conceivable that, under this model, the network's auditing capacity would further centralise into incentivised miners - a relatively small group of participants that has already massively shrinked in recent years. This does not set a particularly positive precedent in my opinion.

Would you care to comment on this concern?

 

[Peter R]

@Yoghurt114

Thank you for the comments. There's probably five points I'd need to touch on to answer your question thoroughly, but for now I'm just going to focus on two: (1) block size growth has naturally constrained itself to roughly doubling in size each year without pressure from a block size limit (2) schemes to achieve coding gain (e.g., IBLTs) are probably necessary to make producing large blocks affordable.

Natural Block Size Growth--the chart below shows growth in the average block size. Since this a log chart, the slope of the curve represents the percentage growth per year. Until very recently, Q* was to the left of Qmax (refer to charts in my talk) and so the production quota had no economic impact on the equilibrium quantity of block space produced. In other words, had the limit been 2 MB instead, the graph shown below would have very likely been almost identical.

If we project this historical growth rate forward (refer to the region indicated by the red circle above), nothing too scary happens. Even in 2020 (5 years from now), it is quite possible that the average block size would still be under 8 MB even without a hard limit.

Schemes to Achieve Coding Gain--Right now I normally pay about 0.0001 BTC to make a transaction. Let's call it 2 cents per TX. Imagine we experience another growth spurt and the price of 1 bitcoin increases to $10,000. Given today's network propagation impedance (about 7 sec / MB according to my estimates), each transaction would then cost $1.00!

Now, imagine that with improvements to network propagation, we can decrease the propagation impedance by a factor of 100 (say we get 25x through coding gain, and 4x through faster bit rates). This would then drop the price of a transaction from $1.00 to $0.01. At 1 cent per TX, Bitcoin could still be used as an efficient payment network.

 

[catlasshrugged]

I enjoyed Peter's presentation at Scaling Bitcoin so much that I excerpted it for posterity.

 

[dgenr8]

Peter,

The room you addressed sorely needed to hear your work framed this way - in terms of economics 101. Cryptographers and cypherpunks are much too quick to dismiss empirically proven economic theory as somehow non-scientific, even though it underlies everything in bitcoin and answers questions they either think unanswerable, or try to address with solutions like "flexcap" which you easily showed to be sophomoric.

During the talk, Peter Todd tweeted some kind of spherical-cow insult, and was quite unresponsive to the question "What's your explanation for why fees are paid today?"

Your exposure of the obvious transparent parallels shown in anti-market responses of censorship and attacks, and of market reaction to these, were so helpful. You are that rare guy who seeks the truth, and also is not afraid to confront its malefactors.

I'm just very glad you said all of this in that particular room of people. At least now, when all the obstructionist garbage is long past, they can't claim they were surprised by these ideas.

At the end, you mention the Ledger journal. What is your role in that? I hope it's a significant one.

 

[Peter R]

@dgenr8

Thanks for posting. Reading your comment I feel like my talk communicated everything I hoped that it would!

Since publishing my paper, I have been in awe at the lack of understanding of basic economics by prominent people in the Bitcoin community. It's funny: the same people who preach the "don't roll your own crypto" line, not only attempt to "roll their own economic theory" but completely dismiss a few hundred years of work by smart people thinking really deeply about it!

The other thing that's funny is that the people who seemed (at least to me) the most pro-free-market are now proposing highly interventionist techniques to control the block size (and thus manipulate the economy). If we adopted something like FlexCap, we'd almost surely end up with a committee like the FOMC deciding how to adjust the price of block space to achieve some dual mandate of "decentralization" and "Blockchain access." In other words, we would have just traded the "interest rate" lever controlled by the Fed for a "block space price" lever controlled by Core Devs.

Regarding the journal Ledger, I presently serve as co-managing editor.

 

[Yoghurt114]

Thanks for your response.

Natural Block Size Growth--the chart below shows growth in the average block size. Since this a log chart, the slope of the curve represents the percentage growth per year. Until very recently, Q* was to the left of Qmax (refer to charts in my talk) and so the production quota had no economic impact on the equilibrium quantity of block space produced. In other words, had the limit been 2 MB instead, the graph shown below would have very likely been almost identical.

If we project this historical growth rate forward (refer to the region indicated by the red circle above), nothing too scary happens. Even in 2020 (5 years from now), it is quite possible that the average block size would still be under 8 MB even without a hard limit.

I'm not particularly interested in the way the block size has grown in the past because, as the saying goes, past results are no indication of the future. From the perspective of participating in this network as an independent auditor, I must consider the possibility of the block size growing 1000-fold - or any size which prohibits my ability to validate the network - in the next year or whatever (had there been no limit) likely.

I can think of any number of ways the blockchain may naively (and in my opinion wrongly) be used (rather abused) to solve all manner of real-world problems - had there been no limitation to the block size - which would result in a system I and many others would simply be unable to validate. That, to me, is unacceptable, as it would destroy the utility of Bitcoin as a decentralized system. I would be more than willing to scale up considerably, but there are limits to what is realistic - I believe the result of your paper would end up with block sizes far beyond what such a limit might be.

The current limit, unfortunate as it may be, is what's keeping all this at bay, and somewhat forcing things to explore ways to employ the blockchain in a way it does scale. Factom is an interesting example. As is the proposed Lightning Network.

Schemes to Achieve Coding Gain--Right now I normally pay about 0.0001 BTC to make a transaction. Let's call it 2 cents per TX. Imagine we experience another growth spurt and the price of 1 bitcoin increases to $10,000. Given today's network propagation impedance (about 7 sec / MB according to my estimates), each transaction would then cost $1.00!

Now, imagine that with improvements to network propagation, we can decrease the propagation impedance by a factor of 100 (say we get 25x through coding gain, and 4x through faster bit rates). This would then drop the price of a transaction from $1.00 to $0.01. At 1 cent per TX, Bitcoin could still be used as an efficient payment network.

Things such as IBLT (and sort of also the relay network) don't work in ways like '7 seconds per MB', it works like '7 seconds for the full block - regardless of its size'.

Yes, it will take time for this constant-sized message to propagate through the network, but it does not affect the orphan rate; in these schemes, the size of the block has not anything to do with the size of the message used to communicate the block. The Shannon-Hartley theorem is not violated because these schemes allow us to sort of tell the recipient which previously-communicated payload to 'select' in order to reconstruct the intended message.

Your paper outlines the way block propagation happens in the peer-to-peer network currently, and it would completely hold if it hadn't been for the fact that the contents of a block - transactions - are already known to every participant as they, like blocks, propagate through the network. This last bit allows for more efficient constant-size block announcements such as IBLT to be a reality, and it, frankly, somewhat breaks the result in your paper.

At 1 cent per TX, Bitcoin could still be used as an efficient payment network.

My interest is not with the cost to push a transaction onto the blockchain. My interest is with being able to validate transactions that have been pushed onto it and *know* the system *is* what *the system is*. (I'm fine with scaling up considerably; BIP 101 is somewhat on the 'far too optimistic' side of things, and rather uninspired, if you ask me - but I can roll with it if the desire is great enough (luckily it isn't), albeit under protest.)

At 100x more throughput (as the 100x decrease in propagation impedance you mentioned would result in) we would have an effective block size of some 30MB or so. That's probably fine.

But what if propagation impedance drops to, effectively, 0 - as an IBLT or some such scheme could pull off. What then? Do we have a boundless block size limit? Would transaction fees not be 0 satoshi? And would not a single entity in the world be able to validate all throughput?

----

And as an aside, I would argue against Bitcoin being an efficient *payment* network (now or ever), for reasons unrelated to this debate. We're *using* it as a payment network, but it has become quite clear this has some problems. Bitcoin is a settlement network, clear and simple. The analogy with Bitcoin being a digital form of cash payments sounds nice, but it's false and breaks down right quick: When I give you a dollar bill, I don't need to wait ~10 minutes for you to receive it, nor does the entire world see that transaction.

Waiting for payment is waiting for settlement. Bitcoin is a settlement network. And not even an efficient one at that. But it's secure, independent of anything - decentralized, and *relatively* fast (which isn't hard considering the competition). Payment networks may be built on top, as the Lightning Network proposal demonstrates, but Bitcoin is not one.

 

[sickpig]

@Yoghurt114 this post is a rare gem. Usually posts of people who see the bitcoin as a settlement network rather then a payment one are aggressive, full of ad hominen attacks rather than proper arguments. Kudos to you. Really. That said I mostly agree with all you've written.