BIP101 and 8GB blocks by 2036

[Cryptology]

I would like to focus for a moment on the future implications of BIP101 which in addition to an immediate x8 fold increase by Jan 2016 also mandates a doubling of the maximum block size roughly every 2 years taking it to 8 GB blocks by 2036. The argument of BIP101 proponents is that increases in broadband and storage capacity will compensate for this additional full node workload. Is this a realistic assumption to make for every current a future node?

If BIP101 is implemented it is going to be a huge gamble for Bitcoin. Putting a high strain on the network will certainly mean that the lower end hardware running on low speed connections will have to drop off the network. Only nodes running mid to high end hardware on fast connections will be able to keep up. Up to which point will increase in transactions and block size will promote full node concentration? I don't have an answer and I haven't seen any study that attempts to model that. One thing is certain though. Centralization and concentration of power can only harm a cryptocurrency whose value is driven by the breadth and strength of the people accepting it as a currency (a.k.a. community).

Personally I see block size increase as something potentially unavoidable in the not so distant future but it should be done with a more gradual and conservative approach taking into consideration that Bitcoin is a complex system and any significant change carry a degree of risk and uncertainty. The more radical the changes the bigger the risk and uncertainty. Coincidentally enough bitcoin price is hitting a 6 month low right now as the global financial markets are taking a hit. Bitcoin price should be surging right now not plunging.

[1715030182]

1715030182

[Soros Shorts]

Running a node has never been free. There was always a cost but so far people have just eaten up the cost. If bigger blocks come too fast what I think will happen is that we will see less nodes run by monetary freedom enthusiasts and more datacenter-style nodes run by for-profit organizations. I hope advances in technology can blunt the effects of this, but we definitely need to find another dimension to scale up in addition to simply increasing blocksize indefinitely.

[Carlton Banks]

Running a node has never been free. There was always a cost but so far people have just eaten up the cost. If bigger blocks come too fast what I think will happen is that we will see less nodes run by monetary freedom enthusiasts and more datacenter-style nodes run by for-profit organizations. I hope advances in technology can blunt the effects of this, but we definitely need to find another dimension to scale up in addition to simply increasing blocksize indefinitely.

It's essentially a staged version of how Skype was transformed from a pure p2p network into what it is today; Microsoft's proprietary spy camera network.

XT is stage one.

[jubalix]

Why do they so aggressively double....why not make it a slower ramp up over a longer time

is the BIP101, without the XT IP address ban thing?

[Cryptology]

Why do they so aggressively double....why not make it a slower ramp up over a longer time

That's the way BIP101 was conceived: https://github.com/bitcoin/bips/blob/master/bip-0101.mediawiki

[Carlton Banks]

is the BIP101, without the XT IP address ban thing?

I'm not convinced about the IP ban thing, and I'm not pro-XT. It's a whitelist in XT, not a blacklist. Still not good. But "IP ban" actually is FUD

Why do they so aggressively double....why not make it a slower ramp up over a longer time

BIP101 is an exponential growth scheme. Exponential growth is unsustainable. The idea is to force p2p supporters off the network over the length of the exponential curve, absorbing the increasing costs as you go, happy that you are exchanging your subsidy for commensurately increasing dropout rate of the p2p supporters. Once this "node running costs" curve reaches a steep enough portion of it's arc, the various types of changes that consolidate oligopoly into monopoly can be introduced.

Shame on the XT cheerleaders. To your eternal shame.

[jubalix]

is the BIP101, without the XT IP address ban thing?

I'm not convinced about the IP ban thing, and I'm not pro-XT. It's a whitelist in XT, not a blacklist. Still not good. But "IP ban" actually is FUD

Why do they so aggressively double....why not make it a slower ramp up over a longer time

BIP101 is an exponential growth scheme. Exponential growth is unsustainable. The idea is to force p2p supporters off the network over the length of the exponential curve, absorbing the increasing costs as you go, happy that you are exchanging your subsidy for commensurately increasing dropout rate of the p2p supporters. Once this "node running costs" curve reaches a steep enough portion of it's arc, the various types of changes that consolidate oligopoly into monopoly can be introduced.

Shame on the XT cheerleaders. To your eternal shame.

Exponential Growth at that rate IS unsustainable. Look at chip speed and transistors per unit, and also bandwidth, it just has not happened that fast. HD space maybe, but its sfae to scale at a slower rate.

I would like to hear some views from in the know proponents of the particular doubling rate chosen.

I think a rate of every 2 every 4~6 years seems a bit more reasonable, and even then that seems high.

The reasoning being that BTC needs to not alienate many of its mining / user base to quickly by centralization.

It would seem logical at least to have doubling or whatever occur in line with block halveing.

The rate of the block halveing gives some idea of how satoshi saw the time scale of things and the time frame to implement changes

[Carlton Banks]

Exponential Growth at that rate IS unsustainable. Look at chip speed and transistors per unit, and also bandwidth, it just has not happened that fast. HD space maybe, but its sfae to scale at a slower rate.

Well, some of those technologies are running into dead ends with their previous growth paradigm (transistors as you say), but they're coming up with new and/or lateral growth paradigms within the same medium instead. But in the end, silicon as a substrate will be the hindrance to progress.

To give you an idea though, the 3D NAND cell stacking technology that Samsung are leading the field with currently. They ran out of road: trying to get even MLC cells to last longer than 10 years or so is difficult at below 20nm pitch, so they stopped digging. They've moved back up to 32nm, stacked in up to 48 layers in the brand new Samsung product, and likely won't use anything too far below 32nm for NAND flash until the layering technology is much more mature.

Another would be the HBM memory technology as a new avenue to boost GPU performance. The previous paradigm there was shaders per die area on the chip (long pre-dating Intel moving CPU's into having areas of the logic circuits on the chip die that are repeated multiple times), but graphics card manufacturers use the same pitch in silicon lithography as Intel currently are, and so really it's more been a case of how many graphics cards you can bridge in single machine (dictated by power and PCI Express bus width restraints). Now the arena of improvement is being moved closer to where the action is happening; the interface between the GPU and it's memory, although what they're trying to achieve is to physically position the RAM in such close proximity to the GPU that the interface barely consists of anything at all. So the evolution of this paradigm presumably will be that the GPU and the RAM will eventually become a single manufacturing unit. The technology uses something fairly new on the market called "through-silicon-vias" (TSV), which I think may have been around a few years, but never commercialised up to now)

[jubalix]

Exponential Growth at that rate IS unsustainable. Look at chip speed and transistors per unit, and also bandwidth, it just has not happened that fast. HD space maybe, but its sfae to scale at a slower rate.

Well, some of those technologies are running into dead ends with their previous growth paradigm (transistors as you say), but they're coming up with new and/or lateral growth paradigms within the same medium instead. But in the end, silicon as a substrate will be the hindrance to progress.

To give you an idea though, the 3D NAND cell stacking technology that Samsung are leading the field with currently. They ran out of road: trying to get even MLC cells to last longer than 10 years or so is difficult at below 20nm pitch, so they stopped digging. They've moved back up to 32nm, stacked in up to 48 layers in the brand new Samsung product, and likely won't use anything too far below 32nm for NAND flash until the layering technology is much more mature.

Another would be the HBM memory technology as a new avenue to boost GPU performance. The previous paradigm there was shaders per die area on the chip (long pre-dating Intel moving CPU's into having areas of the logic circuits on the chip die that are repeated multiple times), but graphics card manufacturers use the same pitch in silicon lithography as Intel currently are, and so really it's more been a case of how many graphics cards you can bridge in single machine (dictated by power and PCI Express bus width restraints). Now the arena of improvement is being moved closer to where the action is happening; the interface between the GPU and it's memory, although what they're trying to achieve is to physically position the RAM in such close proximity to the GPU that the interface barely consists of anything at all. So the evolution of this paradigm presumably will be that the GPU and the RAM will eventually become a single manufacturing unit. The technology uses something fairly new on the market called "through-silicon-vias" (TSV), which I think may have been around a few years, but never commercialised up to now)

Very detailed and thanks

BTC relies upon several parts of tech, the slowest dev one will be the bottleneck.

It would seem that the communication side would be the slowest considering that fiber optic / wifi or etc is largely dependent on the gov of the day, the country, and legislation as well as many vested interests. You physically have to put a lot of stuff in the ground on poles etc and then be prepared to update it.

In many countires this simply is not happening at any where near a doubling rate every 2 years.

There will be new techs for this I'm sure but even then incumbents get regulations passed to make it unlawful to use or supply so they can make their profit. The NBN tesltra/Optus/TPG /ASD in Australia is a classic example of where competition is stifled.

[DarkHyudrA]

What worries me more than HD/RAM/GPU when the blocks are alowed to be this big, it's the network "weigth".
I doubt that everyone will have access to a ridicously good ISP with a ridicously good internet speed like Google Fiber.
This will lead to a major centralization to only the very riches around the world. Even light weight clients would not be that light weigth, because running a node will be too much expensive, and then using Bitcoin would mean PAYING to have a simple wallet.

[glub0x]

BIP 101 is really not exponential. It stops at 8GB.
Much more like an S pattern if anything.

The real exponential BIP i know is the 17.7%/year...

[jubalix]

BIP 101 is really not exponential. It stops at 8GB.
Much more like an S pattern if anything.

The real exponential BIP i know is the 17.7%/year...

see why don't they just make a slower run up

[Muhammed Zakir]

BIP 101 is really not exponential. It stops at 8GB.
Much more like an S pattern if anything.

The real exponential BIP i know is the 17.7%/year...

Honestly, are you just saying this BS to show you support BIP 101? Did you at least did a calculation before saying so?

[glub0x]

Honestly, are you just saying this BS to show you support BIP 101? Did you at least did a calculation before saying so?

Yep it looks pretty clear to me BIP 101 stops at 8GB certainly not exponential.
It has some exponential growth phase just like loads of things in nature but then it settle. Noo drama here...

[coinpr0n]

Honestly, are you just saying this BS to show you support BIP 101? Did you at least did a calculation before saying so?

Yep it looks pretty clear to me BIP 101 stops at 8GB certainly not exponential.
It has some exponential growth phase just like loads of things in nature but then it settle. Noo drama here...

It's growth has a sudden halt (at 8GB) but it is exponential. And looks nothing like an S-curve. Really basic maths.

[glub0x]

All the time i read that "Exponential Growth at that rate IS unsustainable. " But the fact that it has an exponential growth is really not important

If the growth was +400MB/year until 8gb,  after 20 years we would be also be at 8GB, would that be better for you? Surely not.

The exponential Word just add useless fud leading some to believe it will double forever (which would be much more problematic)

The problem is the rate and the cap, not the exponential part.

[sgbett]

Honestly, are you just saying this BS to show you support BIP 101? Did you at least did a calculation before saying so?

Yep it looks pretty clear to me BIP 101 stops at 8GB certainly not exponential.
It has some exponential growth phase just like loads of things in nature but then it settle. Noo drama here...

It's growth has a sudden halt (at 8GB) but it is exponential. And looks nothing like an S-curve. Really basic maths.

The argument is that it is unsustainable, which relies on the exponential growth being unbounded. The growth is bounded therefore the premise it is unsustainable is invalid.

Really basic logic.

[Carlton Banks]

All the time i read that "Exponential Growth at that rate IS unsustainable. " But the fact that it has an exponential growth is really not important

If the growth was +400MB/year until 8gb,  after 20 years we would be also be at 8GB, would that be better for you? Surely not.

The exponential Word just add useless fud leading some to believe it will double forever (which would be much more problematic)

The problem is the rate and the cap, not the exponential part.

the rate

not the absolute value

the rate

go to school

[teknohog]

It's essentially a staged version of how Skype was transformed from a pure p2p network into what it is today; Microsoft's proprietary spy camera network.

Skype was proprietary to begin with. It uses p2p techniques to relieve network congestion, but it was never p2p in the sense of an open-source grassroots movement.

[Carlton Banks]

It's essentially a staged version of how Skype was transformed from a pure p2p network into what it is today; Microsoft's proprietary spy camera network.

Skype was proprietary to begin with. It uses p2p techniques to relieve network congestion, but it was never p2p in the sense of an open-source grassroots movement.

the way I remember it, they used p2p to grow the network from zero, but added the supernode structure once they got established.

Agree it was proprietary code, but I meant proprietary holistically. code, network, standards etc

[Klestin]

Once again, for those who missed it the last 100 times: BIP-101's exponential growth of block size cap can be adjusted via a soft fork.

[Carlton Banks]

Once again, for those who missed it the last 100 times: BIP-101's exponential growth of block size cap can be adjusted via a soft fork.

Up or down.

[Cryptology]

Once again, for those who missed it the last 100 times: BIP-101's exponential growth of block size cap can be adjusted via a soft fork.

I really miss your point. You mean that a hard fork is necessary to go from 1MB to > 8MB blocks but once there block size can be "adjusted" via a soft fork? if that is possible why is XT a hard fork?

[teknohog]

What worries me more than HD/RAM/GPU when the blocks are alowed to be this big, it's the network "weigth".
I doubt that everyone will have access to a ridicously good ISP with a ridicously good internet speed like Google Fiber.

8 GB per 10 min is about 100 Mbit/s, which is currently available for residential customers in lots of places, using VDSL2 or cable for example. Of course you'll have more traffic in practice, but that gives a reasonable order of magnitude. By the time of such block size, technology will have improved even further. Besides, the numbers are upper limits, the blocks need not be filled to the brim, but provide enough capacity for the rush hours.

As others have already said, running a node has never been free. Personally, I've found it's the advances in mining that have made Bitcoin more centralized, i.e. less attractive for the hobbyist developer. For example, when the first GPU miners came about in 2010, people including Satoshi talked about breaking a gentlemanly agreement. Mining was obviously more centralized to those with more hardware and money, but it was still fair game for all. Soon we were developing the first FPGA miners which was great fun. However, the next stage was ASICs which leaves little room for the humble developer. You just buy the device from some faceless company and use it -- or if you have more money, you buy more of them and further centralize things.

Nevertheless, it all has made the network stronger for everyone, and everyone is still free to use it. I guess many others feel the same basic dilemma about lots of things: should it stay as a fun hobby project, or should it grow into a serious business?

[dserrano5]

8 GB per 10 min is about 100 Mbit/s

But you don't want to take 10min to transmit the block. You want to do it in as little time as possible, some dozens of seconds at most.