How a floating blocksize limit inevitably leads towards centralization

[wtfvanity]

Some of you guys are really overthinking this.

I hardly think so you're not talking to me.

[solex]

You know what would be almost functionally equivalent. Release bitcoin2 as soon as it becomes a problem. it could be literally exactly the same as bitcoin1 in every way except 2. The market would drive people to adopt bitcoin 2 and 3 and 4 when transaction fees became unresonable, but they would also have incentive not to addopt too early since the more established chains would be more marketable...

Anon136, the problem is that while competing alt-currencies are perfectly acceptable within the cryptographic community, it would be a major pain to the public at large.

The public just want The Bitcoin. They have no stomach for multiple variations, it would be like the VHS/Betamax or Blu-Ray/HD-DVD debacles all over again, worse, bitcoin 1,2,3,4 would be the same kind of problem on steroids!

If it can all be done in the background with multiple blockchains seamlessly integrated, invisible to the public so that there is only bitcoin - then great. But how soon can that kind of technical solution emerge? Not as quickly as saturated 1Mb blocks will happen.

[conv3rsion]

10a Use cases for Bitcoin decrease as more transactions types become economically not feasible
10b New investment in Bitcoin decreases as potential use cases decrease
10c Additional integration into Bitcoin decreases
10d Network effects decrease value in Bitcoin
10e BTC / USD drops



Ask Friendster what happens when you can't f'in scale. Anybody remember Friendster?

And this is exactly why some of you guys shouldn't be so hostile towards out-of-band transaction systems, even if that results in some degree of localized centralization.  For that matter, member2member transactions on MtGox is localized centralization, but does not imply that MtGox suddenly has some special power over the Bitcoin community.

I'm definitely not hostile towards out-of-band transaction systems. They're GREAT, if you trust the centralized resources that provide them. I'm hostile towards a hard permanent limit of 7 transactions per second / 50GB per year transaction growth in Bitcoin regardless of the growth of storage and bandwidth capabilities because some people believe that fewer total transactions of a less used currency (with then decreasing utility and value) will somehow make themselves more money, perhaps, in the very short term until the system dies.

Buy hey, at least you can run a full node on your wristwatch!

[conv3rsion]

You know what would be almost functionally equivalent. Release bitcoin2 as soon as it becomes a problem. it could be literally exactly the same as bitcoin1 in every way except 2. The market would drive people to adopt bitcoin 2 and 3 and 4 when transaction fees became unresonable, but they would also have incentive not to addopt too early since the more established chains would be more marketable...

Anon136, the problem is that while competing alt-currencies is perfectly acceptable within the cryptographic community, it would be a major pain to the public at large.

The public just want The Bitcoin. They have no stomach for multiple variations, it would be like the VHS/Betamax or Blu-Ray/HD-DVD debacles all over again, but worse, bitcoin 1,2,3,4 would be the same kind of problem but on steroids!

If it can all be done in the background with multiple blockchains seamlessly integrated, invisible to the public so that there is only bitcoin - then great. But how soon can that kind of technical solution emerge? Not as quickly as saturated 1Mb blocks will happen.

And thats why, even if Bitcoin isn't the theoretically the best system for micropayments, it WILL be the system for micropayments. Betamax was technically superior to VHS, didn't matter, it lost. Alt currencies might be fine if you are a crypto nerd, but the public won't accept them. We are having enough of a challenge convincing ordinary people that Bitcoins are money. But when we do, they want them and they use them and THEY BECOME MORE VALUABLE.

Bitcoin's value comes from its function and trust. If it becomes expensive to use Bitcoins, the game is over.

[MoonShadow]

Something just occured to me.

Part of the problem here is that, while we want to keep free transactions available within the system, users are accustomed to sending free transactions and getting them processed as fast or nearly as fast as fee paying transactions.  On some level, this is a psycological problem, and one that we can address fairly easily.

What needs to be done, is alter the transaction priority scores so that, not just spammy transactions, but any free transaction will not realisticly be included into any block for a time period.  If we were to alter said priority scores so that all simple transactions that were free (or with a fee lower than the minimum) had to wait two or three hours before they could be included into any block, then the bitcoin user base would be better conditioned to expect and adapt to such delays if the hard limit were ever to force the issue.  We might actually see some changes in behavior, and perhaps better fee paying compliance as opposed to simply complaints from the userbase accustomed to free transactions.

After all, processing these transactions may be very low cost, but they are not free.  If there is no downside to sending mass numbers of free transactions, there is no rational reason for most people to ever start paying them.  However, if we were to tell everyone that this is going to start happening as a matter of course in, say, three months; Satoshi Dice and others will have the time to adapt their models and behaviors, and we might actually start seeing more people paying the minimum fee out of time preferences before we get to the point that we are bumping up against the hard limit.

Free transactions, as they are currently performed, are charity.  The last thing we really want to do is get an entire economy accustomed to such a charity, no matter how nominal that cost might be.

[jgarzik]

Bitcoin's value comes from its function and trust. If it becomes expensive to use Bitcoins, the game is over.

The market doesn't work like that.  The free market involves parties reaching an equilibrium.  "too expensive, therefore collapse" is not a realistic scenario depicting an open market adjusting to costs.

[conv3rsion]

Something just occured to me.

Free transactions, as they are currently performed, are charity.  The last thing we really want to do is get an entire economy accustomed to such a charity, no matter how nominal that cost might be.

Bitcoin definitely doesn't need free transactions to be successful. But it needs transaction costs on the order of around a cent  (to prevent spam), not a $15 wire fee. What is the actual cost to store a single transaction forever? Some very small (and decreasing) number.

[conv3rsion]

Bitcoin's value comes from its function and trust. If it becomes expensive to use Bitcoins, the game is over.

The market doesn't work like that.  The free market involves parties reaching an equilibrium.  "too expensive, therefore collapse" is not a realistic scenario depicting an open market adjusting to costs.

We know that the equilibrium when not restricted by block size is minimal fees per transaction (even free transactions) because that is the reality now.

An imposed sub optimal block size does not create a free market, it restricts one.

[solex]

...
Free transactions, as they are currently performed, are charity.  The last thing we really want to do is get an entire economy accustomed to such a charity, no matter how nominal that cost might be.

Actually, free transactions do earn money. They encourage new users. More people will try bitcoin, accelerating its growth. By implication this contributes a proportion to its exchange rate against fiat. So, miners that include free transactions are indirectly benefiting from them as their block reward is higher in fiat terms.

[MoonShadow]

I'm definitely not hostile towards out-of-band transaction systems. They're GREAT, if you trust the centralized resources that provide them. I'm hostile towards a hard permanent limit of 7 transactions per second / 50GB per year transaction growth in Bitcoin regardless of the growth of storage and bandwidth capabilities because some people believe that fewer total transactions of a less used currency (with then decreasing utility and value) will somehow make themselves more money, perhaps, in the very short term until the system dies.

Buy hey, at least you can run a full node on your wristwatch!

IIRC, that 7 tpm metric is a calculated max rate based upon the current average transaction size and the 250KB soft limit.  The hard limit is currently one MB, so we can presume that 28 tps or so is the realistic limit.  This doesn't sound like much, and it's not really compared to payment processors such as Paypal.  However, as a payment processing backend network, comparable to ACH, it's not really bad and it's a couple orders of magnitude cheaper.

The real question is, does the main bitcoin network really need to support such transaction rates?  I would argue that it does not, and that I'd give even odds that a one MB hard limit wouldn't result in catastrophy.  I would also argue that a 10 MB hard limit, which would put us up close to paypal, isn't a bad plan either.  Keep in mind though, as we raise the limits, the theoretical resource usage by the network goes up exponetially.  We need out-of-band networks in order to fulfil Bitcoin's original promises.  They have always been expected to develop around the main network as the economy grew, we just can't really predict how it's all going to play out.  Free markets are like that, efficient because they are unpredictable.

And if you can run a full node on your wristwatch I'll eat my hat.

[Anon136]

You know what would be almost functionally equivalent. Release bitcoin2 as soon as it becomes a problem. it could be literally exactly the same as bitcoin1 in every way except 2. The market would drive people to adopt bitcoin 2 and 3 and 4 when transaction fees became unresonable, but they would also have incentive not to addopt too early since the more established chains would be more marketable...

Anon136, the problem is that while competing alt-currencies is perfectly acceptable within the cryptographic community, it would be a major pain to the public at large.

The public just want The Bitcoin. They have no stomach for multiple variations, it would be like the VHS/Betamax or Blu-Ray/HD-DVD debacles all over again, but worse, bitcoin 1,2,3,4 would be the same kind of problem but on steroids!

If it can all be done in the background with multiple blockchains seamlessly integrated, invisible to the public so that there is only bitcoin - then great. But how soon can that kind of technical solution emerge? Not as quickly as saturated 1Mb blocks will happen.

Yea it would take time.

[MoonShadow]

...
Free transactions, as they are currently performed, are charity.  The last thing we really want to do is get an entire economy accustomed to such a charity, no matter how nominal that cost might be.

Actually, free transactions do earn money. They encourage new users. More people will try bitcoin, accelerating its growth. By implication this contributes a proportion to its exchange rate against fiat. So, miners that include free transactions are indirectly benefiting from them as their block reward is higher in fiat terms.

The exchange rate doesn't earn the network any money, what you are describing is an externality.  And it's a bad policy to get new users, normally accustomed to a three day wait for cheques or a 30 day wait for CC's, accustomed to an hour wait for free transactions.  Yes, we should have free transactions, for many reasons besides your's.  It is not necessary for those free transactions to finalize at a nominally equal rate to fee paying transactions.  If you are donating to a charity, you can wait for three hours or three days; after all, it's a gift.  If you are buying something from Wal-Mart, WM can suck it up and sponsor their own mining company if the wait times are problematic for them.  But if you want your transaction to be processed quickly, you should get used to paying for that.  Users should get used to paying for speed now, before we are all forced into it.

[conv3rsion]

...
Free transactions, as they are currently performed, are charity.  The last thing we really want to do is get an entire economy accustomed to such a charity, no matter how nominal that cost might be.

Actually, free transactions do earn money. They encourage new users. More people will try bitcoin, accelerating its growth. By implication this contributes a proportion to its exchange rate against fiat. So, miners that include free transactions are indirectly benefiting from them as their block reward is higher in fiat terms.

THIS RIGHT HERE

Every person that converts his USD to BTC to play SatoshiDice (at minimal fees) accelerates its growth and therefore its exchange rate against fiat.

Transactions shouldn't be free, they should cost what the market dictates they should could. A market that doesn't limit them based on an arbitrary technical restriction that fails to scale forever with computing resources.

[solex]

"Actually, free transactions do earn money. They encourage new users. More people will try bitcoin, accelerating its growth. By implication this contributes a proportion to its exchange rate against fiat. So, miners that include free transactions are indirectly benefiting from them as their block reward is higher in fiat terms."

I am not saying if this is good or bad. I am describing the reality of the situation that exists.

Has no one heard of a loss leader? http://en.wikipedia.org/wiki/Loss_leader
Bitcoin is exponentially growing against fiat currencies and established payment systems. By accident or design some/lots of this growth is via its loss leader of free transactions. Arguably, SD abuses this facility.

However, loss leading might be the best "winning" strategy that will have BTC kill off most of fiat in the shortest time (if that is the goal). When the competition is decimated all transactions can have compulsory fees (if that is the goal). Bitcoin does not exist in a vacuum and obtaining value in fiat during the growth phase is sensibly leveraging an external. The network benefits as most participant nodes would have a long-term BTC holding.

Hotmail and Yahoo mail, Google search, Facebook social network, Huffpost news: all are free, all are loss leading to get market-share! Once/if they decimate their respective competition their fees will rise. Their goal is transparent. In the vernacular: to monetize eyeballs. Loss leading is essential to get there.

[dooglus]

I'm pretty sure that the 250KB limit has never been broken to date.

block 187901 499,245 bytes
block 191652 499,254 bytes
block 182862 499,261 bytes
block 192961 499,262 bytes
block 194270 499,273 bytes

These are the biggest 5 blocks up to the checkpoint at block 216,116.  Interesting that they're all 499,2xx bytes long, as if whoever is mining them doesn't want to get too close to 500,000 bytes long.

[MoonShadow]

When the competition is decimated all transactions can have compulsory fees.

I don't ever want to see compulsory fees.  That would undermine a great deal of transactions in meatspace that we don't want to see.

As the system is designed, there is no way for the receiver to pay for processing.  There are conditions wherein it's ideal that the burden of such be upon the merchant.  The Wal-Mart example is a general one. 

If Wal-Mart starts accepting BTC in their brick & morter stores, most of the time they are not going to need confirmations for their funds.  However, when it comes time to pay their employees, they need to have those free transactions processed.  So it's in Wal-Mart own interest to support a mining operation, if only to garrantee themselves timely transaction processing.  The same logic applies to Target or McDonalds.  This is what we want.  We desire that the big players be willing to support the network, even at a loss, much like a bank must buy a quality safe.  The safe is a cost center for a bank, which mostly deals in electronic forms of transactions anyway; but no one wants a bank branch without a safe.  Such a form of off-network compensation for mining would establish a baseline network infrastructure, and thus result in more 'sticky' difficulty.  Would more marginal miners be pushed beyond profitablility?  Probably so, but the network ends up with more reliable infrastructure and a lower security cost overall; and Wal-Mart wont do it unless they figure that it would be cheaper for them to sponsor miners to include all of the free transactions that they receive than to expect their customers to pay the minimum fee per transaction.

[hazek]

Someone comment on this: https://bitcointalk.org/index.php?topic=145754.msg1551072#msg1551072

[solex]

I'm pretty sure that the 250KB limit has never been broken to date.

block 187901 499,245 bytes
block 191652 499,254 bytes
block 182862 499,261 bytes
block 192961 499,262 bytes
block 194270 499,273 bytes

These are the biggest 5 blocks up to the checkpoint at block 216,116.  Interesting that they're all 499,2xx bytes long, as if whoever is mining them doesn't want to get too close to 500,000 bytes long.

I understand that at least one miner has their own soft-limit, probably Eligius and probably at 500kb.

[MoonShadow]

To make a point about this, I went onto BitcoinWatch and checked the transactions for the past 24 hours.

61,689

On a per second rate, that's about 0.714 transactions per second.  While it's true that we've hit some pretty high highs.  Has there ever been a case where we were maxing out the soft limit for a full day?  Or even for a full hour?

I'd like to see what a transactions per day chart might look like over the past month or so.  I'd guess that it's not likely to be over one transaction per second.  While 7 tps sounds like it's slow, it's a continuous process, 24 hours per day, 7 days per week.  That comes out to be approximately 4,233,600 transactions per week.  Think about how often you use a credit card.  Do you use it everyday?  While this number isn't remotely what it would have to be to handle the kind of transactions normally done in cash, we don't really want to handle every single transaction.  What if we look at it like it's a bank account.  Normally, people get paid and deposit into their accounts once each week, and then they widthdraw spending cash and sometimes use debit cards, etc. to buy things or pay regular bills.  So if we were to grant the average user 5 transactions per week, we'd currently be able to manage this kind of rate to just under a million users.  So if our goal is to have as many direct users of the blockchain as there are citizen of the United States, we'd have to have a blocksize limit of about 75 megabytes.  However, a blocksize limit of merely 10 MB would manage roughly 40 million users who directly access the blockchain on a regular basis.  This is the practical equivilant of 40 million banks, if out-of-band transactions are both practical and encouraged.  That does not sound like centralization to me.

Keep in mind that your full client functions like a bank for your single account, but an online wallet service does the same thing for an entire group of members.  We really want to encourage out-of-band transactions if we are ever to achieve the kind of success Bitcoin is actually capable of, not to mention that out-of-band transactions improve anonimity if done the right way.

[conv3rsion]

While this number isn't remotely what it would have to be to handle the kind of transactions normally done in cash, we don't really want to handle every single transaction.

Why not? Because it would require 25GB blocks (to do 15,000,000,000 transactions per day) and that seems completely impossible right now? Computing resources are growing much faster than either the number of people on this earth, or their number of financial transactions. Think about how valuable Bitcoin (and the resources that protect it) would become if it ever got to that scale. I'm not saying this is going to happen tomorrow or even in 10 years. But in my mind, more (paying) transactions = more success for Bitcoin.