Hashing It Out
Hashing It Out

Episode 33 · 2 years ago

Hashing It Out #33: Loom Network / Plasma Cash - Georgios Konstantopoulos

ABOUT THIS EPISODE

Plasma is one of the strongest candidates for Layer 2 scaling solutions. Since the Plasma whitepaper was released in August 2017, there's been a multitude of Plasma implementations experimenting with the idea. On this episode we feature Georgios Kosntantoploulos, Lead Researcher at Loom Network, to talk about Plasma, the different variations, and the future of scaling through Layer 2 solutions.

https://github.com/loomnetwork/plasma-paperhttps://www.gakonst.com/

Now intery cinetwork welcome to hashing it out APOGASP forWesolkd to the ATTECH intevators behind Blockin, intrastructure anddecentralize networks. We dive into the weeds toget at why and how people buildthis technology, the problems they face along the way ime listening and learnfrom the best in the business. You can join their renks all right, guys, episode. Thirty threeof hashing it out today is an episode. I think about a topic. We've beentrying to get for a long time, but first as always, call in Cuche, hello,Collim, hello, Colin, Hello, calling hello,Colin. Yes, Ye h, Ho Guess sppos to say, say so:I'm actually doing Odeman like a operiting, but, like you didn't do that.You just wanted me to repeat it extra MEDA, no way fair enough, er right. Our guest todayis Georgias C Concantopalus. I hope I said that correctly at this time:Plasner Research and development at the loom network, which is a project that Iadmire quite a bit based on the work they're putting forward and some oflike the brand new tech of if theyr network, um on't, you say, what's up, tell us howyou got enlor Ha as well as my Um, what you do and what the Luminup work is right. So, in Loom, I'm building plazmaa specific ga on the plusma called PLUSMA cash, and we want to use itbecause we're building infrastracture, amongother things, that involves building scalable blockchans and we believe thatPlazma is currently the best solution towards trust, minimized, scalable,Blok, trins that are anchored to other blotins for their security and what the hell is a Plasa, no work.What is that at? Like Ogool, ten thousand foot viewof plasma for the Lidin Lso before getting into Plama? Maybelet's discuss. Let's give a refor review of Wat e side chain es, and thenwe can say about how Plazma bilds onside Jens and what it improves soI'll go with. Sidins is basically a technique which was aliproposed, Ibelieve in twenty fourteen by Blockin in others, and it involves the tacticof you lock up in Assid on one block twin and you have another bloc trainwhich effectively verifies that you locked up and ass it on ther, first onewhich we call the main chain. If you will- and this means whenever when itverifies that the acit was locked on the mainthain it creates, it means thesame amount of that Aset on the side crain and you can use the acit in theSid Tin for as long as you want for whatever you want and when you want towithdraw the acid back to the original chain. What you're doing is that youburn the Asid that you had on the Sidedran and Youan Lok, the asce thatyou initially looked up in the main scene. So an is way to think of it. Itit's lockup, mint burn and lock it's the two step closes the cappens. Thisis a very nice process, it's very good to use it for interperability so thatwe have two block: Tims Wei move and ased from bit going to a thery O myconduct, maybe an ICO on sum with Bitcon, which is two way packed. So anyany bicoin that I put on a tm its I. It still means that it is one bit going,even though that lives on a therum. However, it has, it does not exactlycomplement toward scalability. Why? Because, if we use, if we connect two blocktimgs that are individually secure, fine, it works. However, it doesn'twork fora scale ability, because if yo use two proof of Wot blook thingsthey're, not exactly scalable, we want to go over something that is morestalable, maybe pfor authority, maybe DPOS, maybe whatever you want, maybeyou want it to be knowsome, Cendralis, Databas doesn't matter. So what we dois that if you have a sidecam which doesn't have a secure, Consenso, AlgoIth, what can happen is that the SID Cain is a Castodion of your funds. Soif you have your side, if you have, if you move your funds on the side, Traineand decideting decides to sensor you and does not allow you to take yourfunds out. Your father are stuck so we're trying to solve to find a way tosolve this, and this is where plasma comes in so in Plazma. Essentially,even if you get censored at any point in time, the side Tran, theClasmachaine, we call it make...

...commitment, t the maintin about it,sait and you can start withdrawal of your acid to the maintrain byspacifiing. The latest wilic state that you know is possible. So the maindifference here is that plasmine is essentially an non Castodia side treambecause in normal surgin cestodial, in the sense that if it faims your fundsare gone. But in this case your fands are safe at any time. Okay, that's a reasonably efficient,ten thousand word view, or ten thousand foot view of what's going on here. Okay, so why has there been so many differentnames and changes and things going on across the PLAZAF infamentations overtime is: I is it because, like you've run into issues you didn't know, you'dhave l e d. If so, what are those issues or are there other things thathave come into play? As, like the initial from from the initial whitepaper of H, like justip ton, an Betalic Verstu' like where we're currently atnow, 'cause Thase, wo things are very to and Farito me yeah. So, let's justto refrain that question to get to the answer to that question was what wasthe original proposal like? Yes, O I tote din, Reh, so moitinally, it's such a look that was known butvitalic had been the post in a blok. I think blong to Tetrom, where in twThwsan and fourteen or O thosand and fifteen whenhe ca talks about theconcept called Sado chains, mmand, it's very similar to what was later proposethis plasmine in thosnd and seventeen. So what happened in the classicPlassmas? We call it. It is that in the classic plasma paper from August withJosewonand Bitalic, they write about. They lay out the vision, essentiallywhat you have a chain that makes commitments in the original chain and,U Leveat an ex Gane in order to dlow your funds, and you can also constracttrees of came so it plasmatine OFFA plasmotin of a plasma team, Phe,potentially levered out infinite cailability right. So this was theoriginal thing, so carrenly naughing, that happens, is related to the paperand Papewa make importannt in laying down trevision of what we are e yeah.So the thing is that was, it was very complex because it caus very useful bigning blocks forthe mental model, TAK which would approach. However, things have evoledsince then so yeah. So what happened was that after afew months, Thar was opposed by vetalic again on EFH research. So usuallywitalic is the one who starts. He starts a raple and then we see howit evolves. So what happened was that he wrote a post about plasman WP,minimum viable Plazma, which was essentially a utx o based plasmatainone level just for payments. No Ston contract, no Nestin blockthings, likewhateina paper, was discribing. I it was, I mean Ha Viam product and what itis. It's singly a two inpouts to outputs utoblotting and that's it. It's pretty DSIMPLE. It's a design, anda few months later, a first implementation came out byDavid, not and UH. That inclimentation eventually was built upon on a concept called more viable plazma,and the issue was that the initial PLAZMA NBP it had certaininefficiencies regarding the ways you were withdrawing your funds in case thecontent Omechanis of the Plama Tanpo mylisses, and so the more viable plasMIAs, an attem IPPES, an attempt. It is a solution to thes small issues, andthere is one big problem, though the case is that all the all these both ofthese plasmovarians they in order to function safely, they required amechanism called mass exits. Why is that? Because um any time theconsersion meca of the Sidesa of the PLASMATINE, it can simply make a hugeaccit. So whenever I say exit, I mean an attempt to withdraw your money fromthe plasmacine to to the mainnate, and what the plasmatine can do is that itcan say: Ok, I'm taking all the funds from the plasmochain and Im exitingthem to the Mainnat, and this means, like all users. They effectively needto exit their funds as well before the initial malitia's exit happens and thitis what I mean by as Exin. So every user needs to have to make thinaftransactions so that they get their funds out safely, and this happensbecause and next it it has the so called exit perial the dispute period.And so when I make an exit, I need to wait some days until I can finalize myecit until my exit can be withdrawn and...

...if my exit becomes Chalente. So bysilence, we mean proof that we mean that some proof that contradict theoriginal statement. So I make an exit and the Chilence cancells my exit, soall users effectively need to exit beforeth before the the Coune Partothat was staying eit and currently there is no efficient solution to masaxens, because you would need to compress a information about onemillion up ose in a very small amount of data. We all F this Nata Ome to putit to Chaine to the chain in order to verify it to start the exit, and so in March of twenty eighteen. I think itwas during the time of if se, the immune confeents whathappened was atvitalic, made the talk on a virient of Clasmin ader viren flaw not calledplasma cash, and so by the way, this is the Vilin, the Plasna that I have beenworking on for the last a few months. Okay, so let's rewind a little bit here, becauseour audice is kind of a mixed background. So we have people from UXdesigners to hardcore protocol engineers, Um and Um. Some of them know what plasma is verywell and some of them do not. So if I may I'd like to explain in as few words as possible,what plasma kind of is a little more M, so you mentioned theconcept of a side shape. The idea of Plazma is that you can usea smart contract to stake some Eth, for instance or token or whatever mechanism for for forvalue Um capture there is into the contract and that contract wouldbasically act like you said in the side chain example, as sort of like the lockUm and burn sort of thing, but instead of actually burning the assets, what itdoes is it enables you to lock the acits in there trustlessly h in thetrustless single source of truth, man, tat whatever and then another chaincanthen inherit the value that is locked in that smart contract. And whatthat enables is another chain Lit Shou, say another ethereum chain, forinstance, to to operate on the knowledge thatthat smart contract up there is locked until proof, t a of exit is submittedand that Prov of exit can occur within that second layer lair to plasma chin,and what this ennables is that on the main net. We have this huge volume oftransactions that we're trying to push through, and if we only have one pipe,then we have a limitation on the amount of transactions that can xbe executed.Not Everything needs to be executed within the context of the main chane.So the idea behind Plasmak was to improve scalability by building a layer,two solution on top of the main network, which enables people to Hook or stake value into a secondary chain andexecute you know: Auditble long transactiontrails on Alayer two solution within a certaincontext of that layer. Two Solution H so, for instance, your corporate I' is probably tbad example that you'rcorporate Um your corporate transaction mechanism, so intortally in yourcorporation. You might need your own sort of way of transferring assetsaround internally through finances through budgeting Yadadyata, but youdon't want to do all that on the main net plus there's a privacy issueinvolved in that. So why not move it into your own sort of like sidechaineUm, because SI chains have thatvulnerability that you said before, vitalic Wa decided to kind of like O,extend that concept where you don't actually have to burn or or even removethe asset from the original chain, and you don't have to depend on thesecondary chain. You can actually use proof mechanisms, cryptographic, proofmechanisms to stake in the original chain and exit from the original chain.Any time you want given a certain period like you said it might take daysto do a mass exit, for instance Um. Just that's the foundational level here,but the secondary side effect of that is that once you have this layer twochain, it's also inheritable of the properties of themainnuts. So, even though you can stake in the mainnut, you can also havesubchains below that, and this can allow you to create any multitude oflayers of chains that you can disect the transactional throughput deeply.Instead of going through this one long,...

...horizontal pipe of transactions. I knowthat's a rewind a little bit, but I feel like that. I kind of would clarifyit for a certain particular your rvaudience. Is that an accurateway of describing what plasma's goals are? It is ICURATE. However, I want toinsist that the nested three of James Structure that you, I think that's whatyou closed up with it- is currently not possible so creating a plasmatine of alasmatain in order to far the insegment and like only how very specifictransaction spare blockin disgintly not possible and we're not exactly sure ifit will ever be possible. So That's interesting, that's one of the mainselling points for me on this, and so it's kind of as I just working withparticular custers. I know that's kind of a need within ther within theirorganization, unless we could get like about two billion transaction throughput on a single bloc chain. It's it's going to be kindof what they're looking for. So I'm curious. What are the barriers to that? So why do you have to next? You cancreate multipue class machines, so the current limitation, yesinterepetability between plasmatines, we haven't solved it and it wouldprobably require that you have one smart contract, onatherium and ohavemultiple plasmachines and they all speak to the same spart contruct,because what happens is that if you try to make a transaction on one train fromone than to another, and then you try to wax it this funds to try to make adumb spend the other tain must be able to challenge in some way. So eacs chainmust be aware of each other and youwuld. Do that with a share smart contract oncan, however, whet they just said it's totally theodetical and as long asthere is no either like proper dogument on it or any implementation, I thinkthat it shouldn't be the actual focussing for focus for right now. Iassume, but like the reason that I I feel like that', that was actually abig selling point. Um is actually privacy Um and in Organis in in inorganization aninterorganizational privacy as been kind of like an issue, and so I sawthis is as as it was proposed, a Lo low hanging, fruit solution, but it seemslike it was a lot harder than it was Unisoan what was proposed. So if anorganization wants a gatekeeper between exiting funds to the main chain, thatwould be like your CFO or accounting office or whatever, but they would liketo allocate that information to other organizations and within theirorganization or their cort sortium of organizations and prevent them fromseeing what each other necessarily do but operate on the only the chain thatthey care about. I see what you're saying it would benice. However, I cannot give you a techinial solution to that that it ma,so you need the main point of the nested a strature. It is that each thedeepest level it acts like a court system, but, as you go up, you get upto the sub superime court and effectively your security is as good asa supreme cold. The main change. That is the whole goal of this. I cannot give you a Technina,explaation Caus, I don't think it exist. I think I think it an interesting wayto carry this argument for or the discussion Ford, that kind ofelucidates. Why or where we can go or what we can currently doing is a topicthat was brought up during th DEFCON PLASIC discussion a couple of monthsago, and that is like what are these specific? Like there there? It hasn'tbeen any standardized specific requirements of a plasma chain and whan.It means to be a PLACI chain, and I know that you're interested insomething like this is earnin work towards this, in terms of like, likeformal definitions and requirements of what it means to be a plasme chain ndand how that operates so Um ten thousand feet olerd you. I believethat what qualifies s a plasmo chain as a Plazma, let's say system, it requiresone smart, countract, O etherium and a block Tam or, let's say data base,because it doesn't exactly have to be ablock in a data base. Let's say abloting that tings its data back to the maintain that TAT end nd, an x game andan incentive compatible, a game through which you can take your assets thatsomehow you got in the plasmantine out back to the maintain and so Um. What I tried to do in the last fewmonths is to create a short sort of more up to date, Lao paper whichencompasses, let's say more in debth the Phinicians and explaining what ispossible. What is Notcaantlis werless outlining future work so that we canhave a more realistic, both...

...expectations, because Esgolin said themarpret use and the three of cains was the seiling point, but it ended up notbeing the case. So I believe it's important that we, both let's SAstandardize what Um is currently possible and whatqualifies something to be a pleasant ting. However, we also need to be asrealistic as possible and not oversail the capabilities of Plazma all right in nterms of overselling, which shit on naming structure wh. What's going on with all thedifferent names for different varants of snot right right, so without plasma eOlitinal one with ha plasma MBP, we HAV PLAZMA MOR WEP, we're waiting for mostviable PLAZMA. We HAV PLASMACASH THO PLASMA KAS is a Vegum which has uniquecoins and because it's going, I Umiq you can think of it as gas as physicalcars, but unique from each other. Then you have plasma CAS flow which improveson some of the problems of plasmacas by fit fixing this uniqueness betweencoins and then wou have plume, prime, which is an aubition on plasma. CASSMCAplus mcasflow, which utilizes cytoglaphic echnique calledAsaaimulatos, where the are satumulators, they use, prime numbers,and this is where the crime comes from. So I totally believe that, and there isalso a snark based viorence of Plazma, which you call snap snazma whatever youwant. No, I believe that this naming viiant alsothere was another vilium called Lasmik te based on the itallsided as a jokebased on bit cuin forks, but yeah look way, God Os, so ndi believe that thenaming conventional it started out. Okay, but then it got out of controland it has been kind of counterproducting, because you will seethat easy. You do not mean to give your class Oviand a weird name, to give a credibility, and so I believe it's very important toclassify to have a taxonomy wit. They offer the plazma variants and what theycan do. So I will tare to abstain from talking about the names and I will talkabout the Vadians and what defedinsate them H, so yeah think Yiu yeah, so theGardent Virience, how I classify them they're the ones wh- are either basedon a rut chain, enforced nonfonsibility. So let's like break this down, so Wi dosay Info nonfansibility. So when you deposit a con, you get so when I ep, Ihave five wither and I pa and I stake it. I deposit et, as you said, on thesmart condacqeitherium. What this gives me is. It gives me a unique coin, whichis worth exactly five wetherd on Plat on the plasmachine and if I want totransact, I can only pass that going around so by Roosaing, not bynonfonsibility. If that's what I'm saying that it's a unique coind thatjust has like a soulin value, an the Vari is hat tere. What I mean Itad isthat so y, you consider it's basically a nonfongeriable token. It's B exaclyexactly. It is a valum where, when the viriencewher you considerd the coins that he deposet as non fun book and to givethat sort of a physical ti and so people an literally a dollar bill with a cyralnumber on it, you cut thaapular bell and half it's a worthless dollar bill.You can't do that. Like literally I jut Sat Ol Bou EXAC. This is the example. Iwas going to use that you cannot count it because if you can it have it's adollar beles useless. Exactly the improvement on this variant is thatinstead of deposting and getting one five wete bill, what I can do is Ideposeii reten and I get one hundred zero point: Zero, five ether bills orPokans, if you won't, and what I'm doing is that I can make better likemore funsible tmactocs, with with nonfangible coins like in the realworld. Withouve like multiple cons, wo have one euros like Europe, so I have.I can give you like five like mot five heros, five coints of one hero each andit's the same as five euros. So what you're doing is that you essentiallyhave no infangible tokens, but you have fancible Ux, because you can sein youpay with no consible tolkens, but you have fundabl Ux, because thedenominations of the Tokens are Pretty Small O. is it clear I if I C drown? Analegythere t s it'SSO like a conversion rate between Euro and say y Slanikrena, likeyou can, or the US dollar or China, so Inthi in in this? You know the Canadian dollar you can.You can say hey this, not intungible token represents the amount of moneyyou've staked in this chain, but you're translating it to another valueproposition now are those nonfungeable...

...tokens in an of themselves? Yes, theyare so that means that your you, the funs, the GRANU, live of your paymentsthere. It is as bud as the smallest talking that you own. So if you ownfive zero point: Zero One Tokins, the the smallest payment that you're makingis an increment of Zero Ping zero one. I question from there: How do youtransact multiple nfts in a single transaction? All right, so you canthink of it as arranges. So, instead of thinking of it as a unique coin, youcan think that a bunch of toins that are in a row. So if I have ten zeropoints, ze on coins in a row, I can effectively but transfer them sotechnical spaking, Transwir, Wuld be start index COA offset and the offsetis how many coins I own and able to anfer. All in one is theat assuming youhave serialized a number of coins. Yes, yes, it's assurn you'd have a bunch ofcoings in it all, and this is where things ten to get complicated because,let's say have ten coins in a row and for any reason I I transfer the middlecoin. So if you have indexes zero to nine, let's say I transfer coing fiveto some other user. Now what happens th from the OI that hey had like onecontinuous rains? Now I fragmented my reins in two langes right, and thismeans that, in order to make like to transfer the value that I want a, weneed to make two ternsactions thinking about it visually. You can think of itas a numbered Lione and the maximum denomination of the payment that he canmake is the maximum continues. Let's say slots that youown in thisline. So why does hat need to be representedas a noncungible token in Thi in the PLASA chain? Can it just be an itorablenumber, since everything has a a state value in the main chain? Can't you justbe like Yeh subtract, five from my total? Oh so yers, ok, so the questionis the question depends on so this is whatto discibing is the account model? No clearall plasman designs. Currently,what we call plasma, they are Utt, so model based, usually theethere areothers o Ode to be there, the vilum that they just said the nonfantableones they're in the family of PLAZMA varients, which are duty exol based. Soyou have a atraxaction. You have you take sowith some ID? You have someinputs, you have some outfots, that's it for a count based PLAZMA, I'm not anexpert on this. However, I believe that there are some problems regarding dataavailability because M in order to make an account model. You need some partybasically to keep track of all the balances of all the users and it getscomplicated because you need to make it somehow so that when you're making apayment, the Partte, you need to verify that your balance decreased, while theother partis balance increased in order to verify that the payment happened orthe opposite, and so Um techncal spaking, because we rely always on anentitycalled, the operator or the consensal mechanism, but usualy to makeit simpler. To reason about. We talk about the entity as the operator, sothe operator as tbi to censor transaction. They can do what Everheywo wont and IT Turns Out that actually designing an ex game which is e Phician,so it doesn't require a exit. Calen salent silence like it doesn't requiremany steps for salences it. It's not quite there yet. So there are somevariants that are account based that you can look up on if research on it research and they would beprobably a better point for Elephans, because I'm focussed on Thaxol busansOtcho. So I mean I, I would think a erceit model would probably be betterthere. If they wanted to challenge it, then it would. If you have a central operator, whos intent in entire goal isto maintain the truth of the system. Then,if somebody challenges it, then the operators er the owness ofresponsibility- would be on the operator to counter that challenge,that they would be the ones responsible for maintaining the proof. Yes- and thepoint is that, in order to maintain the security of the main chain, we need todesign a system that does not trust the Operatol, because the moment that youstart trusting the operator Youre back to the original side, chag modelRiyoucat, I saing- is it. If the operator does not submit the proof,then you don't trust the then you you side with the person who cantestit I'm not exactly sure about that as theevailability problem would be pushed to the operator- Oh yeah, yea, so, okay,so what you're saying is that I can Beli on the operator for the daye availability is the th. Is that the...

Oran? So I can post contest on the mainchain any time I want, but the ownest of proof is on the operator. Yes, sogetyou cannot assume data Vailabili from Yor operator. There are twoapproaches to this: you either design your system so that the data requiredis as small as possible so that your client can naturally Jak at once andstory forever or for as long as you need it. So I have a coin and I need tokeep the datetae related to the coin. For as long as I own the coin, themoment that I give the coin away, I can discard the DA fine that works. This iswhat we are doing. The alternative is t if the data required, it's too big, youcan add some sort of data, avalability forthe proof, and I believe that thisis also what etrum two point o is the to do with fhroad proofs. This also.Maybe I think that polkabot also doessomething like this with some front proofs. But the point is that if yourely on the operator for frot groofs for Dat availability, you meet also on theway to punish the operator if they do not provide data vailability right soum, and if you are away, if you are e wayto punish the opera for Dadavalability, Mat is also the offerator may begriefed by grief I mean they may be attacked in some ways at cause them tolose whatever they would lose to. Basically, Ivnoperat is honest, but people may still attack theoberator and make it so that they look disones and uh. This boils down to aproblem called what was it um speeced listener for equivalent? Sobasically you cannot say for sure if the other party acted malicious or theysimply so, if they're, maliciously withholding data or, if they're beingattacked, and they are not able to provide the data. So if I can m make this more tangible,again Um the operator, let's just say, facebook decides to throw up a plasmachain. Okay, people hate facebooks, and so they decide that they're going Tainindate face book with so many transactions and exit contestscontesting you know: coultest Ou, W they're going to contest that this thfacebocause being dishonest, even when they're being completely honest, theirfacebook can't handle the availability h of the number of those con. You know,for whatever reason, basically like a dedos attack, I guess you could say umon their system say by a staye doctor who cares right m then facebooks rating, let's just saythey carried in the main chain at Carma rating, which a BAS SA is basically acredit score, would go down if they're not able to fulfill the proof required to do it. Furthermore,let's just say their Internet goes down 'cause. They are all on alet's, O, saythey're on aws and for some reason a we s decided to you know as shut the bedfor like half the country. Well, then, what Woa? You know, what are you goingto do? There's a timeframe and and EFUS had a problem and all their stuff wason the cloud and then suddenly they're they're they're not able to prove that they're actually actinghonestly. That is an attack on the operator which D make their Carma godown and there's no way to reverse that a thatn AACURATE COFE cannot defedenceit if tpe ishonest or their being attacked Okaycoo, and if that happens, yeah theOperado will not provide the service ot. You say you can't rely on operators forthe availability problem. So let's get back to your taxonomy discussion. Thatwould be an account base system, but that obviously has some flaws. Whatother kind of systems are exactly so I mentioned t a rousine enforced onfomthebuilid, then o the the violence with the nonfanbulk tokens. Basically, andbasically, what we're doing is that we try to fix, try to get fundable UXS forthis and the there's one minor issue withthis w. This Vien is that in order because you need to have the full- UTke, O history when you're making a transaction the full past. So if I havea Coinan I'm giving to you, you no to verify the whole coins Ce story, sinceit got imitialtly deposited in the Plamatin to make sure that you're notreceiving a counterfeite coin. So this history can actually get quite big, andthis is why we want to employ some sort of Zeo knowledge proof in order to makethe R contstory more compact, and so, firstly, we have nonfandible co coins.We are fundable Ux to them in order to maintain the security by their onPansibility, but improve the U X. and finally, we hadd some sort ofcomparassion mechanism for the coins history in order to make it even easierto have a light plane. So this is the nontandrovion. The other varian is whatplas might be paying more viable PLA.

Are they are Plazma Ukx, so bas, Plamamodiles, which rely on ordering of exits for security? What does this meanso earlier? I said before that the operator can exit the whole value ofthe classmachine. This means that if the Plasmantin has like ten millioneiere in it, the operator can create out of nowhere one huge duty, Xale and exitthat and how does Ama more viable Plasmas, Olvis Bat. It says that we orether exits based on some priority and essentially what this means is thathonest users, even if they initiate an exit after the Malitio Seit of theoperator, they prove that their excit has higher priority, and this meansthat when the exits get finalized during settlement, all their exits willhappen and they will withdraw their value and the operator when they triedto finalize their exit. They would not have any value left for them towithdraw, and is that clear, I would wind why that we need Ihave, Mor ditingyeah. I understand the reason for it, but how do you quantify that priorityright? So the order in Plazma N, BP, so they're there're, two violence, minimal,viable and more viable in the minimal, viable Palasma. The ordering is basedon the outputs on the age of the ALTOCA. So if you have aneeet on the output formore viable Plama, it's based on the inputs. However, the x game is a bitmore involved and I think there were some reason development that I beinread about it. So I would again refer you to their document on ithresits likefor and Jus, not saying anything, no to say anything wrong, but basicallythey're ordered based on the input. Are you able to? Are you able to holdup before before? I ask that question Um, arther, more variants that we havenot gone through yet yesof course, one so, the final variant of Clasmas, so onthe UTXOL. Let's a base model of PLASMAS are the ones which are usingzeoknowledge proofs to inforce a valid state tansitions, so all the ones thatI said before they required silences, because if I make an excit I might aswell just transfer the coil to some other user and then exit and this myexipt needs to be somehow shalenced by the new owner of the coin and thisessentially it's an invalid state transition. So the TASMA vilence adusesnarks. What they do is that they have the opeator creates. A merical tree is a snark for thecurrent state, a Zi knowledge proof for the current state of the Plamatin, andthey submitted to the to Thi Plasmamart contract on Jane and the PLAOSPANCONTRA. It runs a verifier for the zeolat proof and itchecks that it is a valid state transition from the previous one. Howdoes it do it? It's simply. You can either call it black magic or what itdoes is that it checkes in the snark. It has all this comprassed informationthat they the values the meital tree values. They obey certain rules, so anyany kind of Ansach on the crappen Din't have the proper signature on it, and so this Raom O plasmine gopulizeZon all that proofs. It doesn't require silences. However, it has dataavailability problems and they saw this vile of Lama. Solvs data availabilityproblems by saying that, if, at any time we lick date availability, westart going back in history. So if we are at block ten and Blockthan isunavailable. We go back to Blok, nine heis that Dat available for block nine.No, we go back to block eight and so on until we find a blook that is availableand when we find t a block, they make an action for a new operator, becausewe considerd that the canidoparator who war withholding is malisious, and sothe last operator gets dethroned and then no paretor cans in to play, and this is the final viriant of carentLama, a designed I actually kind o got lost towards themiddle part of EP tone more time. I'm really sorry,but like that's, that's where I I sow, you know simply speaking, you justvalubate state transitions on train without putting all the data requiredfor the state transition on chain, and you do this using the s. You knowledgeproof, because you e you're youre, creating a succent proof that theverifier the smart contract contain, can check and say that. Okay, this is avilen state toositon and if it is a vil...

...state to ansision, it will allow you togo to the next. Yes, and if there is a data availability problem, you justkeep rolling rolling up. You go back in time until ther is data available,that's it so who would be submitting that thatthe chain Eto thecaandoperedor, so the's thertheir a responsibility to check in basically? Yes, exactly Ifen, thoughPareto does not oh yeah. This is a designed requirement in all plasmadesign. The operator always submits the metal root, yet the laidest Blok.Always so. That's another question that I haveabout the plasma design who pays for that. The operito it's part of theirbusiness model gotshop. A good way to think about this- is that the oeator.Obviously we will require fees from the users. This feels they do not have tobe in the protocal. The operitor can be O, know, Netflix and say that if youpay me five dollars per month, I will put your tansactions on plasma. I willallow you to Tansac Complasma, so a good mental model is the thing that onplazma you get sensored by default unless you pay whatever you want,whatever Teeri, to ask you for in order to include your transactions. This islike in the total, an Velsarialman said what would happen so then EFOR, becausethe operator they can simply choose do. I include or not thisguytons actionsand yes, careco yeah and the fees they getpaid out of the protocol. The thing is that the costs for putting the menitalrout on chain they can stuck up yeah. So it's to put one word on the mainstane twenty thousand gas. Maybe you want to put two words: Ter Somemi bloks,or maybe it's like around forty or sixty thousand gas, and if you do thisevery fifteen seconds for a year, this is maybe, depending on the price of WiHa. This can be maybe a hundred thousand, maybe one thousand. If if etgoes to Zeo, who knows the thing? Is that Um, the operation of course for theopenator? They they stuck up, and so it is the responsibility of the operatorto choose an appropriate fee model in order to offsit their coasts. So all right so yeah. I get that thatmakes perfect sense to me Um what I'm still Kindo caught upon n a DK snarkmodel is that they submitted to the smit a proof of the current state tothe chance there bas like yeah. This is a valid state that doesn't necessarilyprove that the operator is being honest to all people participating in theirposmachine correct. All it is. It is because Tiat dony certain transactionsfor wellways, Sosoyou're, correct so LASMA, does not have any kind ofcensorship guarantee. If the operadod wants to censor you,they will censor you and you will never be able to make it an osaction. However,you can always exit. So the whole point is that, even if you get censored, youcan get your money out and you will never touch that ope later again. Okay,that makes sense yeah in honesty, that's kind of the way the world worksanyway. You know you could join twitter, but if tuiter says your account's goneand your count's gone right, eally and these these Pasbi Chans an this isimportant for o s to understand: they're, not th, ther're layer, twosolations they're not supposed to be as public to the degree that you know theorium is in which it'sopanpardicipatory model where anybody can kindo. You know do what you know doanything by the rules you're free to to do anything into censorship per hat. Iwanted to a is just not that. That's that's why I wanted to make sure that Ikinda asked is like what, like the the fundamental trade offs of a plasmachain, is like it's clearly, not theas, generalized and fair and trustless asthe base chain. So in order to go to that lair too, you make some trade offs.Are There Fune upple trade offs? Or do you get to kind of pick and Choese baceon how you operate right? So, let's take them one by one.So, firstly, sensorship resistance. No, there is no, or rather not noe. It's asgood as the open innel wants can potentially be none. However, gettingyour Phunes Ain perfectly. You get the Ame, and in order to maintain securityof your funds, you need to obey and chalence and stuff. This is the same asany type of Laer to solution, because you need to maintain some local proofsabout what happens in the sidechain in the plasmatine in the payment Jannal.If you want like whatever so that, if something goes wrong, you chalence thesame thing happens in blackning network. You need to be online everyevery sooften so that Um, you can ilence if somebody tries to get your money out, OTis to Seiv your money. So the zero...

...compromise on the security of yourfunds and as long as you follow the ruls thei rules- are that you love himevery so much time and you check are my frends. OK, fine! Is there somethingthat I need to do. I challenge wo hers that three, the transactionthroput that you get it is lited by the operegod. This can be assumed that if an operatorbecomes big enough, they can route an atbitrade amount of thonsactont, theFFAS that Yeu pay. They also depend on the OPERATO. Evenoperato decides to give you free fees, fine you're good, if not sure, Yo again you like it. The thingis that you pick your own operator based on the services and and whathappens that usually what I'm guessing will happen. What we do is that Ye willcreate a Plasno Cain. The plasmentting will be part of a bigger, let's sayscheme. So in our case we can Essentiallli get some funds in theplasma team and use this to, let's say: Buy Some Bosset Bak for your cat am orbuy a card or sailacard. So it is it's not strictly just like for you, you,you can held insailor, you should not assume the most it. It will not alwaysbe the most adversariale environment, but even if it is you're still safe,but if it's not if you're opimistic about the security of the environment,things will be super smooth, and that is the whole goal. So H, letmelet me S.I'm sure this has been asked within the community before, but let me let metheory craft for a second Um. What ot an operator list plasmachine whereeverybody Regis who registers and stakes is suddenly a validator on thenetwork and a and submitted to a what is essentially a randown contract tomake them responsible. For you know, you know you submitting proof that thechain state history has been updated properly. Fine do it! What do you win?If you do not Ik something Really Tho? Do you gain Youlayin Computationalovercatons? You gave theelay o the random censorship resistance. He sayyeah. I think it would be more censorship, Resistin, O use. Why don'tuse the Baselaer then, because the baselayer has atransactional cost, that's tied to etheorium and TATWeat F Ave, a transactional cost which is tied to the Lassmachaine. I I. I still believe that it's betterto use the tenting soorwhatever, like basically wo L, I would eve, go as faras say use bit going in order to get the most ECENTAL SIP resistance. Thereis like a spectrum here right, so you need to optimize for whatever you need,and the thing is that no oper Atal Will Boer Censoring r trade of a simple likegaming gard operation. If Yours, if your transact, if you're transactingone millionaie IR, which is an operation which may be good to censor,why are you ven transacting that type of mount on the Plasmatine, so I'massuming that plasma, changs or any type of layer to somution tit's, notreally for, like all the huge amounts of transfer that you need them to becensorship resistant? Maybe maybe that's a bit handwavy ifyou will, but I I can live with it yeah I got Ya and I understand like I, ifyou're building a value proposition against a car trading game and andwhich is a great example or Crypto Kitty's another decent example. You know, then, why would you want to harm your networkby by basically censoring people who are Um just trying to do regular normaltransactions on the network? That stuff will be evident to the network? I wouldassume in some way so would everybody be able to see the transactions that intheory, I guess that would be up to the operator whether or not their rulesallow for that, so that people coan know if, if there's censorship going onor like this is. This is the kind of thing where I could see there, possiblybeing a business case where you don't like your competitor using your system.So you start censoring them, and people would want to know that you're doingthat yeah, definitely and also you can always expect that so we're always trying to optimize forlife clients. But you can also assume that there will be nods, which have the same storads asthe Operato, but do not, but do not have right access to the Blostin, so Yecansassume that there will be data availability providers. I am veryconfident, but there will be data vailability to providers giving youaccess to the data so that you know if somebody's being censored or not Gotcha and then theethere'd be probablythird party systems out there just to just monitor that sort of stuff yeahintevoer on it and that be I own business, modeling N of ITSLC. Butalways we must desie. So that is t e important, but we must design thesystem so that it works without them...

...and if it works without them, then fine,maybe teto will bid the build them. But you can never assume that people thatthis MECA will exist. So it happens in the event that the operator loseesstate availability. Is that force Samass exit or is that up to the thechoice of the contract? IY stakeit? So if OPEITO doesn't provide dataavailability, Youd check your local storents. If you do not have the mentalproofs, their proof required to do any action on you, yeah you're done,because you always must maintain the proofs required to prove ownership ofyour data you ar so. It is essentially like running a full node. Just for thecolins that you wown, but nobody else, let's say, has inor about this Gons. So if you lose this data, nobody else hasit Katcha, okay, that is, that is the mostadverciral case. But again I am Preaty, confident that there will be let's saylike so to be here, the current ilimentation that they are you anytime,that you donload that you get a con. It cases all the merical proofs requiredfor it locally. Any time hen you blockcoms. It gets again the data, soeven if it openitor dies, you can still, you always have all the Dati required.So this must be part of your implementation, Iy Se, yes, ND!Basically, you could just throw up your own operator technically and say: Hey,I'm completely forking this crypto kitties and creating my own crypto. Iabsolutely yes! Well that IG makes it inhuge in tha. It actually diminishesthe value proposition for being an operator. Doesn't it well? No, youcannot actually do that because, currently, currently, if on on animplementation level, you specify one abdress that allows you to as to deposit and to ave to update thestate of the counter to do Somenin the blocks, that is a openrater to go to amultioparator model, Youw do to alter it and I don't think it actuallyeliminate avaliproposition. It even makes the operator that exists moreregoruse, because they're they're gaining money through fees, let's USrone. So let's assume that the open, inter is gaining money through feersand if they know that there's somebody else who is ready to take their spotand take away their sweet fees and they will at any time they will even triveanharder to be a Goveritar yeah. That makes sense so and yeah you want to dothe capture, that's it! So it's basically just the same model we havenow except fe're, extending so that we can have the truth. Mechanisms onMainchain back the models we have now with typical webaps Um and withouthaving all the transactions necessarily go through the pipe of a Webbot sense.Yeah. It's localcumsensus essentially loca at consensos on its chain, andwhenever you want to an then open it, you go back to the LOB I've. AlwaysI've always do this as internets to the witer Internet and Thanchaine being theInternet each of these plas machines being an intronet. Much like you haveyour loal, your local DCP panting out ips for the one you receive from yourISP totally it's it's a similar type model in terms of trust. Take thatanalogy what is the Internet of loom like what? What are you making wh? Howis loom using this and a D and building on it? Why do you need it Rightso? Theplavmatis currently used by loom for we have a battleground game, is onlybazzegound hat. We call it a card game that you can it's a tading card game on Acomo thatyou play on your mobile vice or your computer and the differences that wehave a marketplace and the whole game ran o a bloc Te. So the game doesn'tBrando Own ethero man. It runs on what we call a darb chain, it's a side, chanfor a dout and h. What it does is that there are two ways to deposit. So let'ssay you have a Critozonbi on Minnut N and you want to play with it on the Duchain. What youdo there two ways to deposit it. You either deposited as viewing the doctainas an usual side. Can so you trust the side, chage consensus or you can deposeit through Plaszma. So the whole thing is that we allow users to have to picktheir poiseon basiny. They either go with the normal consensial mechanism,which is DPOS, which gives you some faltolerance sent. Tyfour percent for,like thirty four dishonest, I'm dishonest, FO, O, let Mesay you can either deposit Wia what we calla transfort gateway. So you deposit your funds and then their funs are inthe side chain and they are controlled...

...by DPOS or you can deposed throughplazma and your pland. Your funs are propected wia plasma and this allowsusers to basically pick their own security. So you can either depose it stpos and you can SOE DPOS and Fastin asactions and know, let's say, m nobokkeeping on your end, there's no need to watch out for challenges: Exits, noMalitias, operatos or whatever, but Um, but if they, if they train hoes, forwhatever reason your fund o Sta, which can be bad. If you do not, if you'renot careful, so we provide users with the alternative deposit, be a plasma sothat, even if thefunds, even if the Sitan goes bad, they can still taketheir funds out. However, there is a trade off that they still need to dosome more bookkeeping like keep their proofs. Keepthei challenges, dochallenges, and so on and Ey believe that it's good that we allow both ways,because always the thing is that you need to provide the most CCURalternative, but you also need to provide good. U X, it's a certainly a hard Broden soultand I've always loved KINDOF. Following these days, big based on I ean ill gota lot of Um cloud. Slash awareness during the traction duriels,the educational guides for rags mar contracts an the C. The was Basiytefacto standard, an sending people to learn things, and that was a reallysmart move on Yell's part and then the Movement for actuallybuilding out this laerto technology. An a lot of LHELPING duild out this Ledotechnology in a lot of ways to enable new businessese cases so that we havebetter web ats dealing with Value S has been.I don't know I've. I've always appreciated th stuff yo are doing so.Bravo, I could tell you I could talk for with you for hours, probably aboutPlama still, but h yea. We do have to rap up, but I I've kind of like a usecase to keep popping up in my world Um a lot, and I I'm not sure if plas wasactually the right solution for it and I was kind of hoping- maybe you couldchime in Gon arbitrary checkens. So if I wanted to do off, train off chainplaza transfers for something that disconects from the Internet entirelyfor o a arbitrary period of time and then is able to rejoin the network andcheck in its value exchange. Is that even something that you would suggestclassbof being useful for so um? Can You reflat it? Can you likeshow the usegase one? More time won is garmjus black box black box. Let's just say you have a series ofblack boxes: Okay and these black boxes can hop on and off the Brid, meaning that they can. They can losecondactivity and that's by design Um, and then you want to you want them tobasically interact with each other Um. Well, whatothingthat. I can with anyvalue ad transfer, so it could be cerialized assets. You know anythinglike that Um and they want to basically verify that each one is kinda operatingtruthfully kind, O like a localized concensus, I kind of envision them allbeing on a plasmachane m, the same classman Chan, mind you or several sameplazmachaines SA. They coan verify that that whenthey bump into each other n they're all operating on they're all operatingtruthfully and they can all sort of like exchange assets in a localizedmanner and then hot back on the network when they get conectivity again. No, Iknow that that's that's kind of an abstract case and I can't get too manydetails about that, but, like our plus machines going to work with arbitrary,like offline time period. So if you have an operator which is the onethat's doing the check ends is it is it? Is it possible that something can beoffline for U of time period and then decide to hot back on or do you need toput a regular check in period for anything, that's going to be a you know,sending a proof of exchange to the PLASAN chaige Yeahso Um. I'm not sureif this will answer exactly your uscast question, but the thing is that inplasma you always like in any layer two solution, you always need to comeonline. Every time you have a dispute period. So if you have a Jonel open and some disput happens, you need to countat that dispute within a certain amount of days. The same applau applies forplasma, so if you have some assets on the plasmatain and you go off Ling andyou go flying for longer than with Thisyeu period, you potentially maylose your asset because it its possible. It can be the case that the operator orsome other party colluding with the operator attempts to exit your funswere tocanized us it whatever and you should come before. You should comeonline before the exxperiod WEFOR, the discuperiod, to challenge them. You donot come online Chalen, you're dumb, so...

...there needs to be a way to lot, have avery strict liveness requirement. So there is a limemess requirement. Is Theanswer and the lineless requirement. We can probably be either solve wit by sating high thispe period. So if we have hiseperwith I once a year, for example, we can assume that somebody will look in oncea year. Takif things are good in the alternative case. What we can do isthat we can delegate challenging. So you can have a watchtower so that yournote goes offlying for arbitrary amounts of time, but you have hiredsome other party to salence, for you does it make sense yeah? So, basicallyyou can. You can have somebody t go okay. I know this node's offline, I'mgoing to go an extend its disute period on its behalf, because I have extralworldly knowledge that this Tis node will come back online at at some point.It just had it: it's not xtended Agaer, it's not extending the salence Beriod.It is literally guaranteeed that while the node is offline, if any chilencehappens, if any mylitou action happens, I will take care of it. Gotya. Okay, I Gotya, Oh man, my oneother question, and this is kind of like not necessarily exactly in Nourweal houses. I think our audience would like to know a differentiation between general state channels and plasma. Soif you could, he could briefly t love S. OK, so does payments payments and onlypayments and by payments I mean I transfer of a tocanized asset, so thetogines. Doesn't I don't care it can be whatever. As long as it has some sortof way to express its how you movit around because what happens isgeneraliztd plazma cannot exactly exist because I cann't really have some stateof a Ofwadu Haf, a chest game and accident.An exit is explicitly an atte station that I want to withdrow some value anda why you can be anything ahicorcanized. However, in order to do state channels,or rather do generalize plazmind smart contract on Plazma, I believe thatstate channels and class myself are complimentar. So this is actually whatwill be working. What is going to be in my fontion, the next few months, isthat we can make we can initialize a statechannel with a deposite from PLAZMA. So let's say that I already put my fundson Plas, my danters acting with them, and then I found I opend a state canefrom plazma and when Iwa I do whatever I want in the states Ari place on chessor whatever. And when I closed the state Sael, the funs that were Beth inthe chesst ti go back in the PLAZNA, so espencially, it's not likely one tothree. It's more like layer, one maintain and halvelayer to Clazna Astatchannels, and they speak to each other. That's awesome, and that's probably,where I'm glad you're working on that that'sall aaiyeah and the eotive. It would be to do vm on Plazma what they call Avmin Ivin, which is quite complicated and, in the end it atchieves the same goalbecause essentially you're funding a smart contract from some state ofplasma and you're settling some value back to plasma based on what happenedit, the smart Cotra. Even if you do it on a smarl contract in plasma or if youdo it on a smart contract that was executed through a state chanrel. It isthe same thing. What rertaining is about the result, not how it Capen, and I claim that they way to do it inthat with the state ongals its simpler to do than with Vmiivian that's great and and yeah. I I justwant to point something out th. The purpose of this research is is not toUm necessarily Um, solve all the problems, but to providethe community with a greater tool set of solutions to solve thearchitecture. Issues surrounding building these centrloks biding glokes,right and so what's interesting to me, is you created this taxonomy of typesof plasma chaines and each one has a Nich use case in its own pluses andminuses, and just like, when you're building scale, architecture on a datacenter or building scalable solutions in Software Youare, building, scalablearchitectures in consensus and and value asset transfer, and I feel likethis is the paradime shift that the community should be looking at, because I I,while I do believe that layer, one scaling is, is something that needs tohappen and things like B. I don't want...

...to call a beacon chain anymore. Was itcalled em serenity, spinity like wh? What serenites doing I is is widening asingle pipe. That's like that'slike laying piping or we givh a verycontraryopin on now. So, okay, okay, maybe sharding I've. I've calledit later too in the past, because layeron is the bicon chain and layer tois each Sharg Jain attached to the Bicam Chane, but we lately came to anagreement that you can call. They lay the BICONCEN incollit like layer, zero,thesarchaned, alae one and eats plasmotine old channels that areattached on its chart. That is the Leer too. I agree yeah. I agree with that, but Lair Zeros, the beacon of randomness,is e Bion of random. Is it also supportnaterases, so ecensivly it very similar to the sitin model, because if in theAMA model Inlif the layer, one gets compromised, aiis also compromise,because you cannot also do reliable, salences and exits. So you marrid go ahead of to go back to mypreviouslike question. Can Ey at a operator less, you know plazma chain,that's what Chartz are yeah, that's what I'm saying is yeah yeah, okay,that makes perfect sense and I, like the whole Blair, zero, Approch, 'causeproof. We count it zero and computer science Um Hads in your language, youshould computers, but H, but you know it's Um uh! I it it just. It's t s, that's whatI was real EKINDOF like asking. I guess, Thereis Shar chains are the wholeoperator list. You know side can solution, whereas yeah this is a layertoo solution. I mean you need to inject an element of sort of like trustiness,not trustlessness, but truseeness, meaning that you got to have this kindof like inbetween state of trust nontrust, where there's enough chucksand balances into to secure yourself properly. So you know your value won'tbe lost, but you also still are dependent on the availability ofanother in and the and Um. I don't want to call it th. Thetruthfulness I guess of Ur Suti Censorship is Yeah Trutinus of anoperator, and it's kind of like this kind of like grey area this this thissuperposition of trust versus non trust, Um, and so I just I. I think I thinkthat that's a good good approach, the eters I wantta. I want to wrap thisup with one last question and that it was like what is there something thatyou wanted to talk about, that we didn't get around to asking. The main thing that I wanted to talkabout was how Badbe Lasma naming is, and I believe, wee builtin through thebit, so we didn't ge like properly shit on it. I kind of danced around it so P, Broa to respect everybody's work,because honestly, everybody's putting so much work on this and uh yea, muchlove in the spekn and everything. Now there's no there's, no, the name onpeople to get as long as I I was surpising at losly, I cou useconfused by all the naming myself and like originally when I read the WhitePaper. It was very clear like this is the direction we're going and thenthere was all these bariance of it and I'm like. I can't keep track of thisand you like. Well, you can definitely not keep track of the white paperplasma, so it's better to have it like this way, also like Iagin, what t learnmore about it, like carforces credit, those those videos online to hell. Ican silike leseal my work right here, so I I wrote something I called PlasmaPaper, plasmy CASS, so with t a paper, an PLASMACASS and it essentiallydescribes at ot is what its imitations are, how how the challenges work, how theexits, work and yeah it's a good rait forever. Anybody who wants to do like afull reading on what Lave Mscavalty an Inos Day, it's apotetly Af to bat likeVegine, I dot Sen et in I that's uring all right. So how do people reach outWHA people o? You can find me on titer t g aaonsand Het tweteris good alrigh. I leit you're, also a GAK onst tcom to yeahOIT Y listeners. If you like to subscribe, Wath the button to tell yourfriends an share it with everybody, you know and we'll see you next week anddefinitely check out what the looon guys are doing. They Gat sommotrosingstuff.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (109)