Vitalik Buterin

[Music] [Applause] since we've got happen on that day we actually side we can start forming a queue so we can asset value 50 positions now so then I yes we go faster how does that sound to everyone okay does anybody have questions if you have questions for you thought you could come over here we start so my question is in the Imperium developers like us by using in führer or alchemy or similar nodes for higher well we can have a better system if you're in tune so that we don't need to have centralized nodes for higher yeah it's a very good question so I think a big part of the reason why everyone is still relying on centralized building providers now it is that Imperium 1.0 light clients are honest we just not very good a not here tonight so for example like even just to be doubly formal like apply and it's something like 5 500 bytes every 15 seconds and then you have to say and then the network is not stable so in the etherium - we've been taking my clients extremely seriously so for example the minimum and in down to about 50 and into sinkhole a claim is something like 1500 lights per day which basically means some at any remotely modern system if you're offline data my client you come back online you can pretty much just download a single file within one second again say I think it's synced again and from their group lens of what a lot of resources into making all of the appliance or Food Network's work so you work really efficient received and what we're over edges and like do all of the critters without potential parties okay I've got another mic so if you have a question before the visit there would be questions off there's one gaze your head then I'll come with a mic and you can ask the question questions after the presentation will have to be about the presentation what's how you want to present it at this event two differences going after the event so any general questions all right thanks coming back to Cali we never thought we'd see you again no I'm joking you're Willy Wonka you come out when there's important news so one of the things I want to ask you is little bit about the libre Facebook's new crypto option is anything you guys are thinking about preparing for when we're this is to come I guess first of all from a technical point of view we bread doesn't really introduce anything like there is some language which I do think is will be interesting but otherwise there's no like jaw-dropping shortening or like and bounty transactions for a second roots or knowledge or any of those things right so the reason why people are interested in Iran all in fitting it comes attached to base point of attention with other large set of company is that so potentially like any kind of strategy for how Imperium garyun projects can continue to have prospering that in debt from some in the context of something like this pain is die point if you focused on technical aspect red blotches way to be focused on how can we kind of leverage to that to bring more the adoption so one part of that design for example and designing bridges should actually not be too difficult possibility and would be trying to have developer ecosystem compatibility between them so it for example there's definitely under famous that critter that pretty tiny given just some centralized exchange eath ones how is the sway and the flow gonna be cajoled and convinced to be for eating so ultimately I think the thing that will motivate people to switch the years easy was the one and moved East you in the long term it's just the facts that needs to offer as much more now you can say that River State can with a Jack like faster finality functioning like clients way worse the ability you know Oh what more efficient general-purpose smart contracts of all of those nice things and of course like it will inevitably take time for that to happen because no matter what we do for the first right here of I do to the distance it'll be a much more untested in the First Amendment eath one has been so far which has survived is through multiple years of attempted attacks and successful das attacks and security hardening and all that and all those other things I am alone and that they think I definitely don't expect people to kind of fight back hard against e to because like all of the core ideas like these days and charting have been I'd have recognized a thing I think as being part of ethereal sport wrote down for a long time there's a theory of plastic which is B which is ready and waiting for them and they've already started to express interest in adopting it and adopting the theory on what mean teachers mmm and then how I think the other part of the kind of transition is just like at some point if they eat one system itself is going to be folded into on beef too and that won't be a hard fork and so like you will have people will have to actually install it and so there is obviously the possibility and they get to doing chain if people just decide not to install it on mass but if people do which I am which i think is a little unlikely basically what's going to happen is won't be the hard fort will take the state route of one chain and it's basically just import it into music uses mechanical noise credible it will continue to exist environment will be basically you'll be able to sense transactions that get processed by exactly the same rules but of all happen to be making my way yes the problem all admit that I personally get more distance from working out back some needs to design I guess the trade-off is like do you either create your own network so you get customized for your use case or do you don't on top of something more generic and like if you like from me I move theory outside of you building on top of something more generic it's a much better option because if the area bits off you only think the benefits from be social like networking right look I get that's benefits from it not a theory ecosystem watch means benefits are there lots of other things that are going to get built certainly advantages yeah it already has support for like the ability to is create like different sub networks in which is we're going to need for charting so discover that high only Amin founder and CEO of intelligence my question is I've noticed some ethereal mail people posting online a four parts plan and they market it some of them pretty much aggressively I wanted to ask first if you support that one as this next [Music] graduations vitally for your first step of blockchain in Virginia theory my question is is what is the future of the theory of home and I know you mentioned about scalability there's a level of change trying to scale as fast as as etherion such as a few others that I don't want to mention but I do know that they are trying to catch up today so regarding how much do you think about the fact that helium should be money so about when designing the crypto economic aspects how important and how much thought goes into making in the crew marketer to premium yeah so I guess like first of all there are a lot of an aligned the interests between like having value and etherium watching the executor because of the price of ether drops to zero then ultimately it thank you no matter what the incentives are the incentives are not going to working people with anything so we have to ensure sustainability is [Music] one of the things that you do say if you do this then after this happened or if someone does something than some option I watch that so anyone can break in perform some operation than the rooster contractors understate whatever what the kajuk we've just what our stages of able to do something else so basically it anything involving kind of non trivial amounts of state storage so for example gonna swap cannot be done in Bitcoin last month and it can't even be done in any kind of logical extensions the bitcoins birthday like simplicity plus bi can't be done he assurance contracts that are in for that that are enforceable can't be done so like you could you White House White House has this issue that you can pretty much so this is the thing where but you have to send one transaction then wait 24 hours then send another within 24 hours you can cancel so there's this large class of applications that you can to in if you're employing systems that you can't do more like systems but the reason why to range in weakness is it's the best way to describe this is because terrific weakness like really focuses on the distinction between programming language that support like dynamic numbers of groups versus programming languages that don't support dynamic numbers of loops and that really does not correspond to this difference between big word scripting and if you remember we're trying to get out so for example like the Viper is thought interactively language it does not support that and what they never have a number of loops but in school team everything interesting that you want to do inside of inside of micro so that's why I together started trying to push for it rich people version of this look what a theory instead of contracting state law in terms of eveness and another thing is that if your email inbox are incomplete anyway because you have a gasps 11 so every program has a hard limit all over steps after what your home city it's all the halting problem so far okay one one last question for this session then you will do some admin and then one starting in ten minutes yeah so I have a question about phase zero and e2 summary understanding is the functionality will be pretty limited but there will be the option to take your eath from each one here one and move them to just yet so as an individual what would be my incentive to do so if I have a limited functionality on keys to basically as you'll be able to do something interesting and you'll be able to earn valid even rewards those are those are basically the benefits of being a validator at least you suspended phase zero and Sophie's euro is that is intended as a kind of warm-up for e2 it's a and we recognize that it will have that it will likely have lower levels of participation than it will in the future but the goal is they're just allowing the chains I'm going to get off the ground running and if they're already issues that appear in a life in a life that we're probably like they didn't have it before there's significant dependence [Music] [Applause] only or three protocols we're here has not heard of channels before your house every channels before so I basically live my channels are kind of two existing big families of layer two protocols that exist as masculine solutions for transactions on block chains today you know there are other things like custodial wallets total side chains for example but I'm trying and of speaking to the trusted purely economic own version so far so things that don't work wire at a specific centralized that third-party trust assumptions so the way that was my aunt's channels both generally work right yes that there are these systems where transactions happen off chain by default and what that means is that transaction data goes off chain and also the commutation involved in executing and verifying transactions happens off chain with three exceptions one of them is people entering the system so people joining a channel or joining a cosmic channel the second is people exiting the end the third is disputes so in this view which basically happens when two different actors in the system disagree about what happened but basically but like because these are fully a kind of Club deterministic systems a dispute really only happens when one of the participants has a bug in their code or decides to try acting malicious way and so forcing the other participants to do something on chain to react in these because of the way in which this weird system is set up you're guaranteed if you have money inside of the layer 2 system you will be able to get your money averaged into the underlying layer level so if you have coins inside of a channel where it's on top water system lowest coins are really sitting inside of a smart contract and the smart contract has logic inside of it that enforces this idea that if you actually legitimately own coins according to the rules of the layer 2 system then you can publish proof data into the contract if you want to and there exists this side game that we call an exit game we are as you will be able to submit data and other people might be able to submit data but if you legitimately oil at own assets that you will be able to get those out sets out and if you do not legitimately own assets you will not be able to launch it when we get those assets out so by default all off kid if someone tries to do something funny then other you might have to react and do some transactions there great feeling lots of scaling so at the same time though right there's there's some limits to what the assembly or two protocols can do so know the Lightning Network like it's been having these capital efficiency issues usability issues plausible Imperium like it seems like we know how to do it for payments but as soon as you try to generalize it to an exchange it gets harder and if you try to generalize it's a really general-purpose computation then it turns out like it seems like without smarts or Starks you can't do it and even with starts or stops it's hard so I wants to try to give it intuition as to why look there is this challenge right what why it's proving to be so difficult to gently or to protocols off the ground and why it seems to be not really feasible to come up we so far come up with a layer 2 protocol that replicates just the etherium experience the ability to write a contract in EDM code publish it to the net and have it be something that anyone can just interact with so this is so I in the in the previous slide I either split up stealing Vita and theta going off chain versus computation going on kid yeah the reason why did that is because the rest of this presentation will be about data via cloth chain but computation stole being off chain now it turns out that a protocols where computation is off chain are not that hard to build protocols where computation and state are watching or harder you know they want to give an intuition for why so this is a protocol for doing just any kind of computation on chain in such a way that the one that the chain finds out what the result of the computation is without the entire computation having to be done on one chain in the normal case right so let's say you have some really complicated computation you wants to do you don't wants to do all of it on chain so if you're technical you want some examples like one example for this might be a purifying a POS aggregate signature with a pile of participants the electrical signature is just one did for a participant but per participant you have to do appearing computation so the computation to be the ratio is very huge so you wants to try to avoid doing the computation you really care specifically about avoiding the computation let's say you have a computation of f of X equals y that you can represent as a series of steps so you take X you pass through F 1 passes through up to thousand through a 3 volt line eventually and why comes out the unrep here is a protocol for it that will interactively evaluating that computation so you have some classic parties that we call some hitters and really something it can be only derision once anyone can just show up and say that and if you submit what you need to do is you need to send a transaction that transaction and that transaction would also contain a transfer of all of the intermediate states right so s in its mediates the S 1 is just X passing through one intermediate state s 2 as act is s 1 and then passing it further through F 2 s 3 would be an asset throughout 3 and so on and so on and so forth all the way up to s n so we take this kind of transcript that contains the start state the unsaved and all these intermediate states publish it in the transaction to chain along with your deposit and then a challenge various notes you know this could be an hour acrimony a week it's purely a security could be just straight off within that challenge period any we can submit a challenge so anyone can try to run the whole computation off cannibal many cells and if they discover oh hey looks like F 1 F or s 1 s 2 all the way up to s 42 or 5 but s 43 is not findable when they get submit 42 as a challenge index and then as part of executing that transaction on Jane f 42 is actually going to be verified ends the chain now finds out that hey the provided escorting 3 actually does not equal to F F 42 of the provided s 42 and so the submitter did something wrong this submitter who is their deposit the Challenger gets party balance or award if there are no successful challenges for a long time then the result is exciting so the key observation here right is that this game is a very friendly to honest participants if you participate in this game dishonest or you're taking a really huge disk and as long as there is one honest participant that's just running through the computation and checking it then it's you know that will honest participants is going to be able to prevent to any bad outputs from getting through it and that one of those producer that's going to make a lot of money doing it so there's lots of incentive so do people to see so being I guess to kind of summarize what what this is doing right basically the idea Syria said we have a computation that we want the chain to evaluate the result because there's some application on Shiamak hears about the results that ieave way of doing it is to just execute F entirely on chain if that's too hard too hard and there's too many scalability issues with that instead we use this interactive protocol and so instead of actually executing F the chain just by default X you just have to download ideas or pull these intermediate states and in general there aren't going to be challenges but even when there are challenges alone needs to run one for one tiny little bit of F so you get a lot of scaling for computation so two people see how a computation can be scaled now data so let's say you have a file and that file has one megabyte if you don't wants to publish the one megabyte the changes is going to be expensive can you prove to a chain that you actually published the file so can you prove to a chain that you publish the file and you're not withholding it turns out that you can't so basically the problems here is that like you cannot have an equivalence protocol for publishing firm if you are applying a billable or data and the reason why you can't have an equivalent protocol for verifying the availability of data is because unlike computation we are once you pre committed to or some value that I used about that is the value you can't change anything afterwards with data availability if you haven't usually didn't publish data at one point in time you can always just publish it at some point in the future so I created this chart that I call the fisherman's dilemma and so the idea here is that imagine if we have an interactive game that's had a similar spirit to the interactive game for computation so basically the idea here is that you have people publish books and if some piece of data in the block is unavailable then you challenge it yeah you say hate watching I think this data is unavailable so let's imagine that you're living in one of these two worlds you're either looking in case wondering you're living in case two in case one piece at time t1 some validator receivable publishes publishes a block of missing data at time t2 someone else have some honest Polly policeman raises an alarm and a time T ok fine we'll just publish the remaining data in case to the original validator is actually published the whole block but have time T to this other validator actually raises a false alarm now the interesting thing here right is that the blockchain is only going to like if it has a way of checking for the availability of data it's really going to only be able to check for the availability of data after time T 3 and the problem is that if you only start looking at the D at that particular piece of data after time T 3 you have no way of determining whether you're living in world 1 or going in role too so you don't know whether the publisher or the person raising an alarm or which one of the two is malicious you know that one of them is malicious because he disagreed with each other but you don't know which one and so thank you habits and well that basically says if you do not penalize making of making alarms in general then if you don't penalize making alarms in general that you're either horrible boss or if you reward making a making alarms and protocol can be bloody pumped and if you penalize making an alarm then basically bad actors can just keep the leg data until no one publishes alarms anymore yes this is just the attackers strategy the attacker strategy is if any reasons of alarms that we just immediately publish the rest of Israel so the point of this strategy is the kind of an attack that just tries to and successfully manage to manages to exhaust that while raising an alarm it's kind of like crit it's basically it's a sort of boy group risible situation right every imaginable fan then every time the wolf disappears and so eventually after a few weeks everyone in the village thinks that the people crying wolf are all crazy and once they stop crying wolf and the wolf actually comes in with everyone so so this is the data availability problem right so basically you just can't scale the now there are availability in the same way you can still combination now there are other ways to get around this problem but they don't fit into the same format of being used usable as a way or two protocols on top of an existing layer one watching so plasma channels lots of ion channels do not solve data availability they bypass it so here's what I mean by this right so here is how plus button channels work as far as data availability goes right so in the a plasma system the philosopher chain operator is supposed to publish all of the data for all of the watch speak into the chain but they don't have to do that they might not do it but fortunately the chain never has to learn whether or not the data is available because if you as a user of the system discover that the data is not is not available to you you just make a personal choice - now the gym so no one needs to hear who was really at fault right basically you just figure up for any particular piece of State inside of the system like to figure out who is there as a beneficiary of that's on that stage they have the ability to access the data and if they're not happy with the chains availability they just active plasma ants channels their exit games both rely on this kind of most recent signature is true the principle right so the we both of these protocols basically work if there is a like if you send me ten coins inside of a channel or inside a logical system then I receive a transaction from you that sends those ten points and in the case of plasma plus oracle branch that proves it that's part of a block now if I were to send a transaction then I have to sign something and intended to send it off to the network so if I have a mortal branch or in the case of channel it's just a signature from you that says you gave me this asset and I personally know that I never saw the message after that transferring the asset for someone else then I have the information that says I have this asset now and I personally know that ability because no one did something there was no operation that could be done later to take this house in a way and so I can publish evidence of their contract and so I can win and get the a cetera right so I have enough information to know that I can win the game and you know basically the whole process just kind of five houses data availability because it's all about well number one the data that you already have ian's number two this interesting fact that you haven't used as a user know that there is no signal there is no signature of sending data from you which is available because you know you never made such a signature so basically both of these game like plasma and channel t-systems really strongly rely on this kind of principle of well let's make the system sits safe by kind of carefully attaching responsibility for state objects to the beneficiaries of their state objects yes a state object for which you are the beneficiary if the data for that state object starts being unavailable then it is your incentive and your responsibility and your ability to start this exit procedure and take it out so there is this kind of underlying assumption that says everything has an odor and furthermore everything has a single owner that in that has the ability to exit that thing from the steps one problem these techniques do not generalize do this one Chris the owner of unis lock oh you know I don't ask you this wall validity shares doesn't mean this one and the person who deployed the unit swap contract theoretically has the ability to vote with all the ether stuck inside of it and some of these voting cows but there's still not the owner doesn't have an owner they violate some of these assumptions I killed these payments the reason why plasma has to do you TXO based payments instead of rolling with a similar account based approach is because plasma info like really strongly leans on this idea that you cannot change the state of someone else's account without their consent that ensures that every one of the Roman state problem with a balanced B system is an imbalance B system you can't muck around with other people's money without their consent just only in the upward direction yeah generally you want to have your money box around with in the upward direction but security during a plasma cash also depends on an assumption that their arms kind of cross dependencies between points so in dependency trees kind of don't really intersect much dependency trees don't expand much as as time goes by historically even atomic swamps bring this and more complex applications break this very quickly so basically the fact that layer 2 system does he exist force you to really explicitly do this thinking about who is the owner whose event she hugged me match incentives here yeah it doesn't generalize to context where the doesn't work well basically being similar to systems have a complicated and radically unfamiliar developer experience and you just can't have this kind of write a contract and publish it feel to be beyond the house this is from Asia blog post and the store Pinta strip who you remember serpent that I wrote back in 2014 you know this is basic this is an idea that I call that shadow chains back then and basically the principle here is that it's a scalability solution that tries to scale called inna or Sri scale computation without scaling data right so the idea here is that once you get published but blocks don't get executed by default will only get executed if someone challenges them so this is amazing you know that I kind of came up with blew it up and at the time I thought that I did not really pursue it because I thought it doesn't really give that much scalability like it's okay you're getting rid of the computation you're not getting rid of the DMX usually gets you're getting rid of half the cost so if you likes improvement in reality that was taller than turned out to be a mistake so it turns out that there is no kind of really nice fact that we can rely on which is that dignity is cheaper than computation and there's a lot of deep reasons why right so first of all like it's just an empirical observation like people have done like analysis of the extent to which bigger or bigger blocks in fact network latency versus the extends to which blocks with more computation gas attack network latency and it turns out that data is currently overpriced and so this is why in the next yeast and Bowl part 4 cough cough network upgrade there is going to be DC guy here's and one of the Yankees is going to lower the cost of call data from 68 yes for white the 16 gas per byte so neither [Applause] second they get a heavy applications or make them or make people watching less vulnerable to adults attacks basically so if you're anyone can remember the 2016 Shanghai gossip why it's always gonna happen is because computation is very heterogeneous there's many kinds of computation and so it's very easy for some types of computation do just perfectly be much less much more difficult for some lines to execute than others and some of those kinds of those attacks are very hard to kind of reality digging on the other hand will data spongeable abide as a bite and so with data it's just it's more difficult to this kind of leakage paralyze the chain by sending the chain data that's formatted and we're pleased because ultimately bytes are just fights so third its transactions getting compressed so in Imperium transaction is about a hundred bytes turns out the majority of that is a signature 65 bytes as a signature you know there's plenty of ways to compress signatures the alleged aggregation is one of them just shoving all the signatures into I CQ snark is another one lots of ways to compress signatures so that to the point where each signature is kind of marginally speaking ends up being blend by or less per transaction second you can compress addresses addresses are 20 bytes aligned but if your entire system has only ten million accounts then you can really complain you only really need 10 million indices and suppressed addresses to free bikes third you don't really need more than one byte for the feet so you can really do a whole bunch of optimization and in many cases you can compress the size of a transaction that you need to put on chain all the way down to ten bytes so and forth we really heavily underpriced the storage and so storage gas prices are going up in Istanbul so Ziqi roll so this is some a different spin on this idea but that works kind of differently so the idea was Iggy roll up is it actually is an instance of this kind of scheme of computation off chain and I beta on Chino model by scalability the way Z control of works is you have a contract and the contract stores the core covers of the stages there are some work over the tree that contains everyone's account balances users send one some network transactions and these transactions are just 11 byte satiric from address to address value failure technically don't actually even need two bits and who would the gods signature then there exist these actors called real ears relating to be provisional if anyone can be a real layer in the Google years and gatherer transactions and we create a CA star and what the CB snart basically says it is is it's a cryptographic proof that proves that if you have whatever the state whose hash is whatever the existing hatches and you apply the signature is that or the transactions that are this set of data so you don't put the signatures on chain just to say just the set of data that's without the signatures just those a lot of mites and if you apply all of those transactions then the state that you get at the end is this new record and also by the way there exists about would balance Englishers for all these transactions I'm not telling you what they are but there exist valid signatures to you you should be happy ability are all officers so you have this cryptographic proves that proves a statement that just says if the state before was this and you apply the transactions that have this data then you get this new state and everything is valid and in the middle then you just publish it on Channel right though we really are just publishes on chain the data without the signatures so ten points per transaction in and also publishes the new hash and it publishes a star the contract takes that if you revise all of it and if it's happy then the contract just sets the roots of even you would provide an intersection so basically the idea here is that you're running a kind of weird kind of quasi sidechain where transaction data is on chin and just actually just enough data is on chain to allow any users and independently compute the entire state of z12 which is important because what that means is that if the existing real layers disappear then you winds can calculate just stayed and he went in started participating but everything else signatures are in the storage so if you kind of yeah 26 what sorry 21 we here thinks it's greater Whittier thinks it's greater than 1% of say 1000 less than a percent right so this is the kind of revamp of the shadow chains idea from the way back and the idea here is basically we're going to do the exact same thing as a key roll up except here we're too lazy to make Bruce Ziqi Starks our public either than hard while address itself we don't want to do that stuff cryptography cryptography is annoying so instead let's do everything we did what with interactive purification instead because high school people can understand interactive a ship where here's a high-school Corsica winter sands how's Edith darks work here is a high school or similar because I can kind of get overactive irrigation works see if there we go yeah this is like interactive verification it's much easier to understand minimal right so basically withdrawals have a delay of have a delay of 100 and blots publish to the chain so when you publish a blog to the chin you do now you don't have to publish as you start you just along with the blog send a deposit and design and that's it right so when I spoke much to the chin can be challenged within one week so the chin doesn't just the store all the latest three patches it also a stores have the hashes of blocks that have been published within the last week and within a week anyone can be challenged and if you challenge the blog then you basically have to provide an index to what the one thing is that you're challenging and you have you also have to provide a witness and the witness basically means you provide to adjust the myrtle branches of the stage that are touched by that particular block and that gives you just enough data and up a lot to be they are applied on chain you never find a block on chain and if it turns out that walk that was published all the all the way back Dennis it with some again Valerie then you revert the entire chain all the way past all the way from that point forward that's it right and then after a week like skip finalized you know so there's we have the Europe live watch themselves if they want to get a pass phenomena guarantee you right so if I pay you to pay you like to die for a coffee and you wants to know right now that the to die are legitimate and I pay you to die inside of an optimistic roll-up like you can just verify the entire off mystic Royal chained yourself and that way you can be happy knowing that because into a blog is invalid there's nobody there to get successful challenge so it's mathematically impossible for a walk together version so movie preserves one other reason why they'll need for starts is amazing is that it makes us much easier that's right you know this doesn't just have to be payment system you could very easily just have fully general-purpose EDM code running inside an optimistic role just a EBM as it is today um what change is that because you don't have a smart you can't just like to make the signatures disappear so you do need to happen to aggregate the LSA make sure and in aggregate be a Lessig make sure is about one extra byte per participant out of that one bit is yeah bit that says go participate or you don't even need that you need the one-bite because basically instant advertised there's a there's a maximum number of participates you can put into a signature for the thing that still be verifiable inside one block and you just divide like of 64 ladies thanks by that oh the signature only gives me reply to some logistics if you're in Gaza limit eight million one o'clock thirteen point three seconds the theoretical maximum the current rule up can do to the map online to buy 759 transactions a second this is a theoretical maximum we've already seen a hundred transaction the second one chain it turns out the main bottleneck isnora gives us not proving so this is like another reason why optimistic roll up as soon as well right so it's got this trade-off where you don't have to deal with all these three issues but on the other hand which you always take a week post instant ball goes up from three to seven fifty nine to a theoretical max of 32:28 so there we go other benefits almost a no on chain state so as he gave a long contract only basically needs to have 32 bytes of state and you can make an optimistic roll of contract also only have 32 bytes of state basically you would just be a requiring in one many challenges this have been extramural riches and you'd have to store more company making the latest rupture easy it doesn't need two environments very robust against future changes like adding brentt's to aetherium like a stateless clients and basically because it's a contract with all what storage it doesn't care about any of that it gets relatively better post welcome so post quantum you we submit of you is a bit of efficiency because instead of smarts you have to have starts instead of aggregate you know what signatures you have to have starts but otherwise needa requirements don't increase whereas any non role on application would have to have basically a huge amount of data pull up in a computation pull off because we have to switch to less efficient force quantum signature schemes also gets relatively better with privacy so the data requirements do go blow up again from its heads back up to about 100 challenge maybe some really smart person zero will figure out a way to optimize and telephone one hash plus ten bytes but avoiding on cape computation becomes more lucrative when the computation you're avoiding is it's just a signature it's also as if you start so that's full unchained data protocols hybrid unchained data protocols so there is this entire design space that's kind of in between roll ups roll ups and things like plasma you can take plasma and you can sacrifice a bit of scalability and gained considerable generality a user experience by publishing some data on the blockchain so for example if you have a plasma decentralized exchange it turns out that if you publish the order matching on chaining then the amount of history you have to store goes down the complexity of the protocol goes down lots of lots of bad things go down and want some nice things to luck um plasma chains that periodically publish history commitments so plasma chains periodically just publishing claims that say for this particular I can on the state object the currents of active state is after this point and if you just publish these claims on that one chain from time to time the amount of history information clients if the story just goes way down the plasma system that publishes data for transactions that go to and from things like you swap that don't have owners but anything between the things that you have owners go watching so this is entire intermediate designs base that could be explored roll ups and charting so when you have shorted watch chains roll ups get even more fun so a lot of basically let's imagine you have the roll-up we and this will be this has to be an optimistic goal became KCK Rolla because this really crucially relies on delayed state calculation so we have a rule where users are allowed to publish transactions on one of multiple charts these then you have a process and this kind of execution game that enforces a rule that says that the transactions get executed to calculate the state you have to look through all the shards and you have to look through them in a particular order first this shard then this drug the MIS chart the next point here then here then here then the next block up up down up down down right now then what we do is that later one we basically stagger the shards or potentially we even allow shard watch to just people wish they ran in for some time and we in Korea have a kind of voluntary measure where a blog producers can just volunteer will be signal in advance and say hey guys I'm planning to publish a blog in C in one second is will send me stuff and so if you have this kind of protocol then even if the block time upon any particular shard is 10 seconds or even one minute then you can have a system where the stage from the point of view of the layer to a rollup system is or be updates more than more often than once every second and so we have kind of de facto leader to always in support times yay sorry could you repeat that last bit please later too old nearly instant block times joy hybrid unchanging some challenges right so probably the biggest challenge here is figuring out a health QR codes work so the challenge of themarket fear is this right so basically let's say that everyone is using roll-up and so we have two orders of magnitude higher scalability yay but let's say you as a user just like you just got some money you have your magical optimistic wallet and then you want to send 0.01 roll of you to pay for a copy well you want to send a transaction probably this that you don't have you don't have any any layer one either right you just have this layer to ether which is this really funny kind of claim inside of this weird groups we can Amin contraction that miners don't understand and so you yourself can't just get your transaction publish to change right in fact when you there's no point that you yourself getting a transaction publish to chain because then you're not going to get all these benefits of aggregation so basically what we need to have this kind of protocol involving intermediaries right you have this network of intermediaries and oh is this recent proposal called gas station network which seems interesting for to create the network of these and these are the media is basically West transactions and collect fees inside of the roll-up yes I would say it's considerably less complex enlightening basically so it's not like a watchtower right because you as a user is do not even subscribe to one you reducer experience as still you publish to an two on two of the decentralized cloud and someone makes it up so it's it has work like the fact that there is a we are gonna mediation at all is the reason why I like thinking this line and saying that it's a challenge but it's definitely not the sort of thing where you have to like deal with multiple kinds of service providers so do all of that stuff so basically you as a user pay your fees and inside of the role of protocol PRP said whoever the relayer is that includes it and the real layer is going to have you from the real layer it's gonna pay a transaction fee including lunch and in Canada minor so the problem here is basically like analyze this market for the risks of this market centralizing if the market does centralizes do risk the censorship resistance one possible argument against is that even if it does centralize that if whoever is the monopoly really gets us start answering people then that just creates a huge incentive basically someone else will be able to collect the fees that are being censored and that someone else be able to just so thing that we're all like we're all actively thinking about at this point okay yeah so two collisions I guess this is a kind of pragmatic moderate scaling paradigm and the reason why I call it pragmatic and water it is because it doesn't claim to offer as much kind of jaw-droppingly high scalability as the all of the various tribes of people claiming over 9,000 transactions for a second with no trade-offs or claiming but the thing actually has fewer trade-offs than other layer two solutions they arguably even a fewer trade-offs than purely run scaling but of course the eventual goal is that you combine this with the earlier one scaling and you have even more even more scalability also there's a big opportunities I'd rather look at this intermediate design space combined option and off chain and on chain techniques to improve plasma and if you have a sharper based later you can achieve it with even stronger properties yay [Music] yay [Applause] so yeah so for the I guess for the second question yes you pretty much have to revert the entire week which is why like if you as a user are receive the payment inside of the system then you probably should actually be replied the internal transactions yourself or have some other mechanism for doing it not to like because otherwise there is a risk that ammonia refers all the way up to a week but the risk is going to be low because she watched us get recruited and everyone Chris submitted the block after that point we show the oyster deposit so you could just say oh it looks like there's been two thousand either worth of the faucets after the story I guess we'll just trust it that's your personal choice for the first question um yes it is or it like it definitely is the case that if two different participants publish like I like calling them packages just published packages at the same time there's like realistically the two packages are going to point to the same parent and then whichever one of them gets included on chain first is going to be valid and accepted and which number one second is just gonna waste a bunch of gas and like that could there could be issues from that it's more severe in CTL and optimistic roll-up because in zq role if you have to spend all of its extra effort generating a cps dark which is actually pretty expensive to generate it's like thousands of times more time more time than it takes a generative block and but even an optimistic role but there's still risk in Gatsby is one solution that I saw is basically you in order for someone to be a real layer they have to deposit youth and then everyone redeposited Eve basically there's a kind of round-robin protocol that just keeps on signing slots in order or possibly randomly assigned slots and so we have a kind of one more to block monopoly during which you can submit things yes good are you concerned at all the racketeering project can say concerned at all about the polka dot might eat use some of the sort of technical implementations that you've outlined tonight I mean what can I put elaboration humanity [Applause] and that's yeah it's fine like we've like you can use our stuff like we've used as in we've benefited quite a bit from one's affinity is so work on BLS signatures for example will we've benefited even from just like watching cosmos watch logic from snake chain than seeing other successes away and what the issues have been I think so I see from a from a kind of business and a customer perspective like a tug of war and so the business wants the least friction where they're you know issuing their tokens or you know using their token in the token economy and the customer also runs an ease friction but they want though at a highest price who tucks most vigorously against the market or like I mean there's a tug of war between buyers and sellers in any market I mean generally I guess you'd expect it to get a competitive equilibrium if the market is competitive and if wonderful signs are less competitive than you had any clubs around but it's just once it's a less efficient and this is regarding a presentation and can you compare them to like the trade on stars for example in lighting this or just because that's what I'm familiar with sure oh so I guess I can compare and contrast quickly four systems one of them yes we are one watching so the second is optimistic roll up there tsuki roll up and the fourth is enlightening Network / cosmos watch channels so the reason why roll ups are better than the later one is because they have about a hundred times more scalability the reason the reason why I think applause Mustique channels and lightning network our potentials are better than roll ups is because they have eventually even more scalability and in the case of channels particularly so including writing including the lightning and all of those you will you get this other particular benefit which is instant instant finality the reason why roll ups are better than plus ion channels and lightning is that they are much easier to generalize so it's much easier to use them for general purpose applications and you learn beyond payments another reason why they're better is that even the use cases that are just for payments tend to cover stray dogs so for example lightning has very high capital lockup costs like there's there's a limit to how much you can transact like you have all the top lightning issues in the case of class well you did like you have fewer of those issues when you have this other issue which is clients needs a story verify history data with roll-ups you double that you don't have any of those issues the main difference between an optimistic vs. hero lock is basically a DQ of the big big big benefits the key one issue is that you can exit instantly instead of exiting after a week you know also like there's no risk of you many of the chain breaks for longer than a week the reason why optimistic role of this better is if that optimistic roll-up first of all is easier to generalize second and easier and easier to film and third potentially has fewer each and other challenges because you don't have this issue of like having a real life a fairly large computation cost that you to generate their first [Applause] you know just wondering hey usually I'm between right now to create system similar to optimistic roller chickypie cream full mode that just down the block there but accepts challenges footing very big transitions so you're saying like can we replicate the benefits of optimistic role of life just massively increasing the etherion change capacity and replacing all of the full moons with my clients and acceptors yes yeah arguably aetherium to just is partially bad right I guess the one main kind of other insight so first of all like the strategy is legitimate because of erm - is adopting it to some extent but also it's not just fraud proofs it's like if you're gonna scale the chain up and you're also gonna need more data and if you have more data on chain with that you also need to have some method of verifying need availability at scale and ability to exist what they do come up with have a couple of extra assumption attached now one thing you could also ask is well why can't we have a blockchain that has like a lower Vedic capacity but gearing like all the computation capacity in I didn't mean it has been explored I mean there's definitely lots of things in this design space to choose from so this is a deployment question so you propose that explained this and great optimistic relevant idea how do we get it deployed and Henry standardized it how do we get bullets to support it how do we get users to be able to make transactions in this way I mean someone just has to build the thing and build it in such a form a foundation sorry so I'm happy to say I kind of understand about 10% of all you say now so my way of processing this is basically we are compressing loading part on chain right and you said the state is not on chain anymore yes so where is all the data um this so the data that is published on chain is just enough data for that users can calculate the full state hopefully if they want to and so there are going to be users that will calculate the state and maintain information to regenerate [Music] yeah I think like it's not there's like a couple of different ways to go about this right look one is that for people who are already writing papers to sign up to more to keep in mind the faction of their like the audience of their papers isn't just people who are oh and almost exactly like them that are as deeply steeped in in the space as they are and they just have to dump and have to dumb things down a bit and it's just part of it and intellectuals job to communicate part so that's like that's one thing another thing is that there definitely is still a niche for people to take existing ideas and debris describe them in a form that is look more accessible than the original form like whenever the original writing is in the case the original writing is not that accessible but no I'd kind of dump down to the point of being like popular science material that says you know it really really is stupid and crazy stuff that's like totally factually incorrect so yeah I mean there's this kind of in-between destructor just could be a captured boar I guess my and I definitely really happy to see more people doing basically year so I know it would just be able to read all of the transactions that went to that particular contract and call that particular function and it would just like run the state-transition function which is process all of in sequence now if the roll-up has been running for many years then like you don't really want to have to scan yours we're gonna watch it so then what you would do is you would just go to the network and you would say hey does anyone have a stage that is more than no so the point is that anyone who doesn't care about the rule of about the optimistic or Ezekiel up system doesn't need to care about it if you are not participating I'm not verifying any of the data associated with this so an archive node of aetherium is not will not be automatically an archive note of every rule upon appearing and if you want super is there is if there is but if you scale too much that all the what you do to make lights clients lighter is undone I would say no because like these techniques don't necessarily as procedurally right so I guess like first of all if you're not participating into one particular layer-2 system then if you don't have to care about it it doesn't affect your security model at all um second I think if you are not participating you are participating in a layer 2 system then they there are going to be waste I have a lunch going for it right so it for example as I mentioned there's this but for if you want to just download branches from the state you can download Merkel bridges you can verify them if you wants to make sure that the state is correct without verifying all of yourself then either you can just rely on the existing deposits or potentially you could like have some federated system where you trust one and then one that servers or you could have a market where other participants just put down more deposits voluntarily so there's lots of ways to do this or that thing okay thank you [Applause] [Music]