村井純のScaling BitcoinのKeynote

public.icon

https://gyazo.com/aa390fa7ad1c4fe17b4676b6961792b1

https://gyazo.com/30f2f00c0182fbe64c15ae77e3cd8512

こんな感じ

見たら書いてるくね？

17 trainers that have joined us from digi lab from chain code from many different companies block stream as well and I'd like everybody to give applause to the trainers of deaf loss plus it would not be possible without you so thank you very much for passing on the knowledge and sharing I'd like to introduce our masters our master ceremony which is indeed Ord Toby hi my name is anteater from DG lab I'll be the MC which generally means I'll be handing off the mics and making the announcements and keeping the time so

your coordination was for the next two days okay so presenting the first top of the day by professor Jin money from keio university okay good morning everybody good morning okay this is Jim Wright from McKay University welcome to our caring birthday oh so let me introduce something about our University I in versity is very old the oldest university in Japan are founded in 1858 so while you I'm pretty sure that you don't have a paper currency I mean right and the highest bill in this country is a 10,000 yen on paper unfortunately and

that this person is a founder of the university so meaning the founding member of the modern Japan right that's that's basically what he did in this country so Meiji era which is after Shokan era edo era and then you know it started in 1868 this is very important part sixteen eight right okay you remember the year and then already working on you know economics society education of course because we are the first one there we are working on an Internet my presentation title is there of course Kaizen is a main title bar I

came from the internet space original so the key investor has been known for the very strong while developer of the internet not only in Japan in Asia globally as well so but then then founder do you know when internet was established found it born in other year when internet was born up 1969 while getting into that ah let me see this one this is a book wrote by our Yukichi Fukuzawa the same person our founder then he visited the United States in 1864 and then he wrote this book and after the cover then the first picture

is a full ocean one family five races brothers all right so the globally this is a you know everybody every human being are in one family okay that's what he wrote even before the modern Japan started Japan was close Shogun era okay so and still it's a 66 two years before major and then steam thing and economics and everything society electricity and delivering communication that time in Japan deliberately was like a person is carrying the things and Iranian he visited the United States in 1864 and he

saw the telecommunication of a telegram thing testing in the 1884 he immediately created this picture amazing see the poles covered the globe and then the carrier is running on top of the wire so this is internal right so he had images already that the world is going to be connected and then they know this is a what we call the Internet pocket right so this is a great vision which we are living in so we call it the internet civilization why civilization civilization when we are talking about civilization then the science and

civilization society and civilization is a very tightly coupled culture and civilization technology of course is the very important source about creating the civilization geography is all meaning you know the Africa Middle East and Asia so geography is important and the segment because of the geography different under the name multiple civilizations in existing together in on the planet and then they know they're for conflict sometimes okay so read these other internet space cyberspace alright cyberspace was created by science

mathematics and other logics Society created on top of the internet culture technology of course supporting the Internet and then the our civilization is a growing by technology Geographic I have a question about this so the our chairman said this is a truly global space Bitcoin is global so then do we care about the geographical difference you are traveling around your country visiting Japan this time segment do we have us multiple internet no internet capitalized single single entity that's what we have implemented so as far as

the you are on the internet and the you are connected you are communicating each other so conflict hope we don't have a conflict but we do have a conflict with existing real space things so that's what I'm told about so before getting to that then let me review the history of our internet this is a 1969 it's a kind of a too important thing about the origin of the internet one is a UNIX operating system of course and the second is apana everybody's talking about the ARPANET all the time but it's a reality is a

UNIX operating system UC Berkeley distribute the tcp/ip therefore global internet exists so a UNIX operating system is really important whatever your favorite operating system is then it's originated on a UNIX operating system because the UNIX operating system was the first operating system for human being users not for the vendors hardware are so started in here 99 and the older history but I have tagged and the blocked are on the each of the segments of the history of Japan are internet right technology is a basically the

basic technology like a tcp/ip and the world wide web 90 and then there I remember when we started the you know kind of a year 90s then the u.s. friends of mine started to do the business and this is IT bubble Silicon Valley everything all my friends went to the business there for that we did the development in Japan so that if you have a Mac OS 10 that certificate protocol stock was created by here ok investing so Apple asked us to donate our source port so we we are we did the reference code for tcp/ip of public PSP version

and so that was the origin of the operating system you are using and so what Japan contributed a lot and then the y2k if you are familiar with y2k my students doesn't understand what y2k is anyway so why do k there is a very strong you know a newspaper articles and everything talking about the internet is going to be destroyed by attacking the thirteen root server operators I mean that means I was chairing the roots of operators community so then if you don't have a DNS and that's a kind of a single

point of failure to destroy the internet that was on a Washington Post newspaper cover story so then they know they asking I've since I was a chair so I was kind of interviewed and no no no we have a redundancy we have a distributed system around the world therefore we don't stop the DNS so that was how we've worked together and the one year later 9/11 and the 9/11 during the US government asking us that I can you can you survive after isolating United States like what we did for the airline Network and I said yes we can survive

that's my answer number one second United States you're not even a die if you isolate from the United States because that you are all did I already deal I your economics to dry on the internet already the third question is it's a technically impossible there could be a way to connect the world because it's a multiple routing algorithm thing so that was the 2001 9/11 time discussion um security it becoming the very important thing and the users by the introduction of the smartphone and everything Bitcoin and

everything so were users having a very strong power SNS of course and the data and the trust probably mixture of all those so those days are basically the very broad view of a history of internet but that tag I put in here it's gonna be very important so what you are living and the we are living under this space so a lot of my friends some of them are familiar according to the history bill is my good friend who started the Sun Microsystems and Jon Postel we work together for the domain name system and RFC do you know

this person anybody knows ok stable is a one of the person to share the history of Japan Internet it might be very interesting Steve was ahead about National Science Foundation I was working for the very first tcp/ip general connectivity working at the NSF office and it the machine was called Mount Fuji and then they know we kind of a bow I'm piggybacking on a cable across the Pacific sponsored by an intensive and then transfer laying IP on top of it and then that time the Japanese government bureaucracies no no no this is we are

living on a telephone work and the telecommunication should be done by a government telephone company and everything you are trying to connect the Internet to the United States Internet is a happening that the military saying and then it's a very dangerous so you shouldn't do that so but we already connected by the way so that they know so they don't approve anything so I said this in the in the NSF Washington DC office and then I told Steve that in in Japan everybody's worrying about you know connecting the

Internet is there connecting the upon that therefore it's a military therefore it's a dangerous and they don't understand so then they you know he left my room for three minutes and he came back with this letter this is Doris no later it says NSF had paper dr. Jim ray in Festival Tokyo in light of our discussion this afternoon on behalf of NSF a pleasure to grant our the Internet access to Japanese communities sincerely Steve of division director so he instantly typed this letter and then he asked me is this help and I said thank

you so it's flame and then I showed to everybody so Internet is okay and so that's what I'm telling all the bureaucracy politicians and the government of Japan I met Steve last year and so you started out the very official our open activities of Internet in Japan and then he was watching this thing he remembered and you know the last year so it was already a trump administration he immediately said if this year I'm gonna be in jail that's what they told me so anyway so that was a very you know complicated process that

the Google internet was a developed under some issues happening in many ways so let me see I'm gonna skip all those story my friend thing okay this this might be interesting the in - I'm sorry in 2006 percent of the world population was so I told you the IP version 6 development Asia was Japan that time and the North America so by the US and Japan working together probably somebody mentioned the very nice thing that the Japanese we going community is good thank you very much and I hope so but at that time

internet Committee in Japan was very much a development side us friend were very much developing the business side six six percent of appropriate world population this year 56 percent now and this was led that was it mm this is plastic plastic plastic actually so us 95 percent Japan 87 percent Europe 85 percent so it doesn't matter over probably 85 percent so everybody using the Internet we are living on the Internet that's why we I call internet civilization eight year by the way only 50 percent yes so 50 percent more people it's gonna

be on the Internet including China including India so that's what we are doing so when you are developing for the Bitcoin scale then this is what you have to see and the thing okay the internet this way so another issue about me I only have a seven minutes to space the internet space real space into the space we are we can do most of the things now and we have a nation in here so we have a nation to nation relations I can tighten nation here so you've been seeing Bitcoin being a global entity you've been seeing all the time with a

local aggression and the other thing we are creating just single space on a cyber space on this panel is actually divided into the nation state therefore internet space is a global truly global we don't care nothing but IP transmission so we can reach everywhere for the recreation different in each of the country right so you might be interested in this right because this is a very in detail i I design see the language goes horizontal horizontal vertical horizontal okay so the police for is China Japan us of course with the

Sun grass and the UK currency different choice I mean legal things are different so then that they have a way to communicate each other right US China might be a trade war whatever the GDP are hates Google and you know so that kind thing is in between the nations right but we do have our space for the global communication so we are living in a boss so I hope the people who'd understand we are now living in a both space so the Bitcoin is working in here and then then you probably that relating to here but the different regulations so

this picture is very important for the anything you do on the internet right so introducing a little bit about Tokyo do I have time five more minutes okay this is a new taxi cab in this town so if you have a taxi waiting for you and if you help find the third taxi in this shape get on this don't don't write on the first one then second one it's not the interesting this is a hybrid LPG electric hybrid car created by Toyota this is going to be a more than 90% of the taxicab in this country by 2020 this is very helpful and the

wheelchair can put in small car but the big space but a lot of batteries in here so talking about the batteries we are working on battery society creating the huge battery safe into the building and you have your harbor smartphone battery and the batteries everywhere and we have been creating this batteries for the laird battery safe and a big finally five years of a development and we finally achieved that so in this case well ac/dc I'm sorry about the language ac/dc conversions the AC DC conversion DC AC

conversion that can be losing their certain percentages were energy so DC supply or by electricity is a really really important for the electric power management right so but they're only happening for the you have an AC adapter for your PC and then you get the AC are converting it to DC and the you move the computer so in this country every single telephone ground station had a used to be a three hours above a 300 the ground station mandatory earthquake in a 2011 electric power nuclear electric power plant bomb

so no power but the three hours everybody's was everywhere everybody's a smartphone was connected and the internet worked for three hours we decided three hours turned out to 24 hours so 24 hours but Lee everywhere in this country store in we call it a tribrid ground station but the anyway the ground stations is equipped with the 24 hours of a tree so we already have a but we distributed around the nation so and DC supply already happening power over ethernet 48 volts and japan t be mandatory this is a TV

show TV trade show and I saw the rj45 huge on the exhibition space so they said connect TV to the ethernet utilizing the rj45 that was a TV trade show for IPTV and the control so mandatory when we this country digitizing the TV then tcp/ip and rj45 became mandatory so not the single TV set in this country without algebra for t5 and the tcp/ip okay so the when you are renting the room your room is equipped with the Internet always always from a star shape from the single point Internet these are getting into a you

house and the other rooms because all the rooms possibly having a TV set right so now the app can supply the DC power in this country you don't need the AC power supply okay so that's basically what we are dreaming about and creating the new battery and the battery is everywhere so we call it the battery society internet based battery society okay now you turn to help us okay you go to the large battery nearby DC to DC charging your battery charging your car and the spending I mean renting the battery from the hospital to your home

okay that's probably the optimal way 400 directed power of that time up I'm sorry but how we can pay for that thing it's easy because LT power you know the evaluation is very easy so this is what we go internet battery society but how we can to taste the micro payment and then name the battery is going to be a worid battery gonna be our entity to for the pay and the al receive and the settlement payment and everything this is the example of what we you know one of the things this country is working on tour since time is

there are two mics on either side and just as last year's scaling you line up and then we'll get to your questions as as much as time allowed allows thanks so the first speaker is Sergei Delgado from the Autonomous University of Barcelona and he'll be telling us about his analysis of dust in you get to based cryptocurrencies thank you thank you very much hi everyone I'm quite flattered to be here today all right I'm quite flag to be here today to told you about the analysis on

the you Jesus had we perform on different you teach always cryptocurrencies and share work with Kristina who is here in the audience and my two advisers from the Autonomous University of Barcelona Jodi Amir sorry all right yeah very okay so let me start by giving a short outline of what I'm going to talk about today basically I'm going to give a brief introduction on UT 0 and the UT goes head really simple I'm sure that most of you may already know all what I'm going to say here then I'm

going to define some of the things we have been doing and how we've been calculating things for our analysis and then later on I will go to results to the most interesting part of thing and give some like easy no conclusions of the top so to start all right well ok so to start well I think that we all know what would you take so it stands for land spent on Jackson help with that it's basically a transaction output from a previous transaction that it hasn't been spend yet so basically when we are talking about

bitcoins we're referring to actual value store by UT UT X owed and all those eggs are stored in the UTC offset which is part of every single Bitcoin full node and we have one entry in the set for every every single of this output so it doesn't matter which is the type of the Osito the murder the owner or actually the conditions at the Scribus encoding and doesn't matter the value we have one single entry for every one of those so some properties of said I would like to point out so this is part of every form

now it doesn't matter if it's born of not and then how these works is that every single entry in the said have a different size and it depends on the UT so type we are actually a I'm sorry so it doesn't matter how much bitcoins were starting in single entry that one actually modified he decided this I sorry but the type of the UT so we'll so our goal for these analysis were basically to see how many and spend output were actually worth spending or in other words how much space we were devoting to store usually exhausted were

not worth spending and we also go out what non were spending means just now so I will talk about two definitions the first one is quite well known I mean it's just well first but we do we mean when we talk about not being able to spend something like something being non non worth it so alright this depends on two main things the first one being how much sighs these itísö is going to contribute when we try to create a new transaction so basically how much size we need to put in that transaction in order to use

this specific output and the second one will be what is the fee rate that the network is expecting at that time and this depends on the transaction below backlog so basically the more transactions in the backlog the more we need to pay right so I as I was saying the first definition I will like to talk about this fast and these are quite well-known definition according by the Bitcoin core this could be has changed quite recently over the last year and this actually used to be slightly different so it used to be was that when

we were checking dust we were referring us an output that need to pay more than one third of its value as fees nowadays from Britain Oh point fifteen onwards if I'm not mistaking disc exchange and now what they are doing is basically to count these eyes and output that actual that need to spend actually more in fees than the value that the output is is holding so this is done all right this is on this way so basically what they are doing it's counting both the inputs and the outputs of the like mmm very

little transaction that we are creating here and the input is always fixed so we can see here this input is like pay to your public key has input and the output is the actual output we were checking so if the field we have to pay times the input plus the output it's bigger than that the actual value of the output then we can see this - otherwise it's not and here we can see we have a value here that exactly reduction use for saying segwayed transactions seguida output sorry so basically if it's a Guidi it is

divided by four so we actually also come in the different definition that was called non profit all and this is way similar now that it used to be but our definition only takes into account the input and let me let me explain this so basically when we are checking the UT Exocet and we see some different outputs we don't know how well we can know how these outputs are going to be spent so we can know which in which are we going to use but we know nothing about which output we're going to create by by

spending vehement so basically our approach was to try to predict as good as possible which input we were going to create and how much size we need to create those and then here the product is basically what the variation I was telling about that will depend to metrics I will show in the next slide and we also counted the same like like segwayed reduction okay so how can we actually know what's what it's going to be the size of the of the input of the transaction so this is just like a genetic interaction

structure in which we can see that there are like some fixed size parameters like the producing text maybe an output and output index or the end segment but then those two are dependent on the on the output type so what we are doing here is basically completely the minimum size by setting a fixed size that will be always four bytes which are the out point three six and four of the end segment and then we have the variable size which as I said will depend strictly on the scripts a and scripts a clang this is normally

between one and two lines so here we have basically two metrics of unprofitability the first one is a lower bound in which we are only using data we can extract from the UTC offset in which basically we just blend with what what is in there and we try to predict only using information from it and then we have an estimation in which we have used also data from like all the blocks in that we have analyzed how these output were in in the blockchain and then we try to approximate as better as possible so then regarding non-sync non segwayed

outputs what we can see here is that for example for pay to public key outputs we have one push of the signature and the signature itself and since we are talking here about like a lower bound we have taken like the most probable small signature since signatures are normally between 71 and 23 bytes and they can be smaller but with a really small parole idiom and we've done more less the same with pay to public key hash and here we can also see how we have a public key that can be either compressed and uncompressed and I said that we have

only used data from the you teacher said this is not completely complete true because here we needed to like decide if the keys were compressed or compressed and what we have done has been to you just like point a certain point block in the invocation and say alright we're going to use compressed from this point on and uncompressed from this point pick by and basically which use like the first block that was mine after the first release of VidCon core in which they were using compressed public keys so quite similar with paid to multi C

and then if we go to pay to script cash you will see that here we have like a zero or nothing here and this is because just by checking page of script hides in the users head you have no clue on how you're going to spend this output so you can not know what was the script that was guys to create this public two pages per public year sorry they paid the script cash this also happens in segwayed well we can see how in beta witness public key hash it's quite similar to be publicly hash with the discount and we have the

same exact problem with beta witnesses clip cars so what we have done here for the lower bound has been to just count the fixed size so since we were not able to actually estimate how this could be in a lower bound manner we just said alright we're just going to take into account the minimal size but the second metric actually has counted these and how we have done it has been by using a block site to get every single every single public key that was used in every single transaction within the block chain and then computing the average

public key size depending on the block so then when we were checking certain util in the set and we wanted to know exactly what not exactly but what was the most probable size of the aisle of this probably we just want you like this block they'd be the average of of size of the block and then put it in the in the exactly theater all right so something similar has been use for page replies in which well as we can see there are like several scripts encoded within page script hash but as I said before fine we cannot know any single

thing about this so what we have done to compute the average input size of a PTO script cache and then use this in the in the estimation and as we can see this also depends on the block size all right so regarding the results as we can see him well here what we have is actually which fraction of the yugioh said can be counted as both dust or non-profitable depending on the metric and we also use a quite wide fee rate because when we were performing this analysis was like at the beginning of this year and well

we are actually now around here so around 25 for those who provide in fees but at that time fees were definitely higher I'm quite sure that you may remember like one year ago that we need to pay around five dollars for one input one output transaction one fees went like crazy high so we wanted to know exactly what would happen if this happen again and try to like cover wider wider range so what we can see here are fewer things the first one is that the dust and the nonprofit automation in blue dotted line and green line actually your

assemble quite good even though they are computed differently and this is because most of the UT exhaust we can find in the set or actually pay to publicly hash and they are really similar when they are completed with both metrics since dust definition of the Bitcoin core use page you public key hash for the input this makes a change and then the actually important thing here what I would like to point out is that if we check around 110 Satoshi's provide onwards we can see how around 50% of the UTS out in the set are actually dust or

non-profitable so what this means is that if we actually go up again to 110 photos you provide half of the things we are storing in the set won't be for me spend then if we actually check which value these Invalides represent so again 100 and satoshis provide we can see how this is almost nothing in terms of like the total between supply at that time what not bit consumed price are a little bit going available at that time so we can see how this is under 0.

05 teen percent of a total Bitcoin available at this time so we were actually wasting like half of the size of the set just to store a really tiny amount of bitcoins and at this point you may saying all right no sorry whatever going to say before so I want to point out that this is nothing new so this has not been having like in the last six months or eight months this has been going on like since the beginning of the coin and we can see how well all those curves has been increasing and this step is definitely more how to say

more well it's more step as steam as time passes and so well what I wanted to say here is that we are reaching a certain point in which this is becoming a real problem you may also give me think that is this actually that bad so we should care that much of this this is actually a lovely problem or not so as I said we perform the analysis not only in Bitcoin but in other cryptocurrencies and I know train like to hit lightning here and I'm not trying to blame anyone here I'm just like showing data but here

we can see the same like evolution analyst is performing lifetime and we can see how like almost 80% of the users in lightning are actually test for almost nothing and this is something that has been going on going on almost into the gaming and the main reason we we found that well we can see that I mean this line is like the 100k block so we can see how this is like this because well the main reason about this is that around 67% of the of all you tick those in the UTC offset of litecoin are worth exactly one little

so I mean this happened at the beginning of the coin it was way easier to to perform this kind of attacks but this can happen nowadays in Bitcoin I mean it's definitely more expensive so you may need someone like with a lot of money and incentives it is but that that can happen so my conclusion on this the first one is that the YouTube though said it's well a big percentage of the exhaust I discovered by test the current implementation of the 80 ghosts of the you cheater said can grow on bandit unbounded so basically this can just go

on and on and on because how is design so it's it's not like who has implemented these guys doing it wrong it's because of how it's how it's based so if we are storing one output for every single voice or it wasn't register for every single output we're creating this would grow and grow and grow and the more the more users we have in the currency it the more the more it will grow then this is especially a problem for low resources devices because as I said you Tito said is part of every

single full node so if you want to like run a full node in a mobile phone as I seen as I've seen like last week some implementation of this going on smartphones or maybe you know Raspberry Pi or so on you'll have a problem because nowadays I think that the users head is around four four in five gigabytes and you should some implementation of it also having an interim so if you try to like have it in ROM and you don't have like enough resources it will be a problem for you but also dust attacks can be performed

and make it grow bigger and bigger and bigger and this can be done by anyone so even though you are actually acting right someone can just like try to make it grow and also conclusions but more talking about like pearls possible solutions of this the only thing I'm actually aware of I'm it's just a proposal it is a TXO commitment by pure thought and I think that also Bram Cohen has like talked about this and I'm not sure if he's actually trying to implement it in like C or not so in current even wrong

so things that every one of us can do definitely output them salvation when when feels all oh so now it's a really a good time for their India's I think that no one of us expected least going so high one year ago so now that they're there are definitely not high we should like try to avoid this before it's a problem problem again and then also whether we like it or not mmm a huge amount of transactions in the net were performed by exchanges so if they didn't use if they don't control it out would use a

good point selection algorithm this can be a huge problem because again I mean it's not just because of us anyone can can make it grow and just to point out this I recall again like six months or so ago how coinbase I used to have like one wallet with 1.5 million you take those storing I mean there was a lot of different in there but not that much for 1.

I'm like what are what are like well I mean I haven't done it but I think that the ones we should all the ones that are like more important for me are the ones that are actually like in the lower part because if we just like put this higher the problem it will be a problem just if we reach these like fee rate in the fuehrer right so for me I think that the most we can do is actually like try to make this go up but the problem is that there's no interesting for doing so and if you try to you try to spend these

these really low low body you take those at the end I you won't be able because it will cost you more that what you actually want I excusing this like at FC last year and some I think it was pure well he told me that if we try to actually like take this this out it will be even worse because like the cost of trying to make this like relate to all the network and all the cost - will cost actually more than controlling this so that way I think that at the a different approach of storing the UT girls that will be actually the solution so try not

to store every single thing just try to get rid of other things - you need like jigs or from communication so when you say output consolidation do you just think about people manually saying I want to consolidate the outputs or do you think about it possibility via a fort of the miners consolidating those outputs by default without a signature from the owner no I'm actually renting like every single user to control your public okay so my question is how do you see an automatic consolidation happening do you think that's possible via Fork

or I mean if I have ten outputs and they are consolidated automatically under the same address or beta script hash without me saying consolidated do you think that's possible or well that will change like the whole thing right if you let all the people like considering your things like just mess with your output without the signature I I don't actually don't see this like happening because how can you decide which will be like that or not how can I just say all right I'm going to consolidate your output

https://www.youtube.com/watch?v=YgtF7psIKWg&t=704s

これめちゃくちゃ面白かった