Hashing It Out
Hashing It Out

Episode 89 · 1 year ago

Hashing It Out #89-Optimism Karl Floersch

ABOUT THIS EPISODE

Karl Floersch is an Developer that is working on layer-2 solutions for Ethereum. Karl walks us through the development Plasma, Casper, and then moving on to Optimistic Roll-ups.

Links:Karl Floersch

The Bitcoin Podcast Social Media

I guessthis ox episode, has brought youby Avalanche Avalanche, solves the biggest challenges facing a theoriumsdeveloper and decentralize finance or defy community. That is velocitysecurity and time to finality, under three seconds on the firstdecentralized network. Resistance to fifty one percent at tax, with completesupport for the etherium virtual machine and all the tools that havefuel defies growth to date, including metamask web three dothas, my etherWallet Remix in many more coming avalanche will be at parity with atheoryum for defi developers that want a much faster network without thescaling issues holding them back, get started today. Building without limitson avalanche by going to chat, dot, avax dot network at is chat, dot a V, ax dot network, xow interywork, welcome to hashing it out tapotcast forretalk to the tech, intevators behind blocked in intrastructure anddecentralized networms. We dive into the weeds figure at why and how peoplefild this technology the problems they face along the way, I'm listening to dlearn from the bestine business. You can join their rakwalkime back toAshnout on Yo host start Oqo, petty today's Cohost Serdine and John, andour guests today is Karforsh, then wanting to get you on the show. I think,since we started this thing, so I'm happy we finally did and uh set startoff with the standard way. Tell us about yourself who you are what you do amazing. I am carflish. I am working on fun things at optimismbefore that I was actually a consensus for a little bit working onUjo and then went from consensus and newjoe onto Atheorian Foundation, working on Casparand the kind of early east to work, and I confess that I got a littledistracted with my eath to work and I ended up getting really involved inPlazma. I was I I M I'm into quick fixes, and so I h just wanted to get it, give right tothat infinity, scale, Um and so worked on PLAZMA. You know a layer, twoscaling technology and then from plazma joined, actually plasma group which Um, basically you know, was working on akind of generalized plazma framework, and you know how do we actually getthis? You know layer to tech to be more general purpose and from there werealize. Oh well, you know a a really really nice way to make it more generalpurposes to you, know, ditch plazma and and go to roll up now now, notablyPlazma is still great, but but that that was kind of the progression, andso we we re kind of disbanded, plazma group and kind ofreformed as optimism, Um and so yea. That's been that's been mylife for the past few years, outside of that pretty normal stuff, bol. So H, I don't have any fiy to ave blanksthere and my audio screwed up for a second. So I restarted again, of course,all right. So so I was thinking so I think themain topic today is likely to be optimistic. Roll ups, Um, probably a good place to start, is foryou to talk a little bit about what roll ups are in general and H.Maybe the backs wrey on the development...

...until now sure Ye. So the kind of concept generalidea of a like roll up has been around for a very long time m. The kind ofearliest tracings of like optimistic rolel up, for instance, was post byvitalic about m shadow chains in two thousand and fifteen, but they'vereally been around for a long time and in fact the thing that has beenprobably the most. That is the most different from you know. Roll ups todayversus the kind of early thoughts about roll ups before was. I think that it isnow more much more common knowledge to the kind of limits and m capabilitiesof the different technologi. So there's there's essentially roll up and there'splasma, and these two are kind of the kind of h, the opposite of each otherin some sense and in another way to say that is a roll ups use on change,transaction data and plazma keeps all of those transactions off chain. So ifa user were to use a plasma chain, they send a transaction to some. You knowsome operator some, you know third party, who is you know, not necessarilyn Elon Minor, and then that party will apply this transaction in some offchain. You know blockchain and then post acommitment to what happened off Chane and by the way. There are new terms forthings like this, like vellidium, which is like the zk roll up flavor of plasma.These names are honestly so confusing. So it's really just like on chain:that's roll up, offchain, Thas, plasma or just say: Onchain data availability,offchaned, it availability, that's like the easiest Um, and so the transaction doesn't go onchain in Plasmo or an offchain date availability, but roll ups. Thetransaction actually does go on chain, and why is this actually useful? Well,to kind of like given intuition well, we want in layer too, to create a blockchain within a block chain in some sense or really a state machine withina state machine, and we want this property that you don't have to sinkthe layer, two state machine, if you're just sinking the lar one state machine,but if you're sinking the layer, one state machine, you want guaranteesabout this layer, two state machine, so ere this layer, two blockchain, Ikindof, use those a little interchangeably, 'cause, they're,'cause, they're kind of similar Um. Now the way that we actually generate the statein layer, too, is by downloading all of these transactions in a role p. Sorollup, we're posting all these transactions on chain and if we 'relayer, one minor, we're just going to run the layer. One Consensus Agam runthe layer, one trn state transition function and we're good. However, ifwe're running a layer one and we want to sink the layr two chain, we will notonly run this layer, one algrathm, wewill Pars, the layer, one and pullout all of the layer. Two transactions apply them to this separate statemachine and sink that as well, and so that gives us a kind of layer, onechain and a layer, two chain Um- and this is what gives a scale, because ifyou are H, you know you, you have the option of sinking Laran tinking, youknow being a light client of Larwan and sinking layer too. You you can playwith what the properties the scale ability properties are of the layer.Two you can play with the trust assumptions of the layer two and it turns out that roll up is a oneof the most um similar to the layer, one in terms ofthe trust assumption. So that's first, that that was the kind of like highlevel. You know what we're trying to do with these layer. One state machines,layer, two stay machines: How these kinds of things you know you can kind of consider them, but nowfor like how the reason why roll up is a little bitdifferent from plazma. The reason like...

...there's a fundamental limit in plasmaso because we keep the transaction data off chain in plasma. We have tointroduce this availability challenge and that basically means that the statecan be Indeterminat for some period of time like one week, and that means that the the programmingmodel, the kind of smart contract programming model, is different,fundamentally different in plasma than it is in roll up in the kind of worstcase scenario, and so this is why we are like okay. We need this like F, weneed transaction data to always be available, so we'll always post it onchain, and we will you know: Um Sink the th O kn w the roll up chainand get approximately the same security guarantees as the layer one m.hopefully that made sense thire's a lot of information right there O. I remember looking about plasma Umand there was a a kind of a lack of fundamental requirements andconstraints to like H or specification in general F forPLAZPA. Has that changed since we've moved thisthis concept over to orlooks like? Is this a general framework that can beapplied to any laier one blockchain under these specific constraints andcircumstances? Or is it or is it kind of Merothe? Ah, it can be applied to any layer. Oneblock chain, both roll ups and plasma, are just fundamental properties ofblockchain architecture and they definitely apply to Wabloc chaings andin fact many block chains use roll ups, but don't call them roll ups, Umthere's. Basically I I've heard I've heard from from nearProtogol to Pocodot a bunch of people. The nearpedal peoplesaid that they were that they're. Basically, you can consider their youknow many Shawers as m. You know many roll up optimistic roll of chains. Imean optimistic roll chains and Um. I I I've looked into or I've heardpeople talk about the pocadot Um, you know architecture and it's similarlymany different, optimistic rowl up Chan. So it's they're all very similar, andreally this similarity comes from the fact that, because layer, one inetherium has tis like general purpose Um, you know turring complete virtualmachine. It allows us to build basically any construction. On top ofthat, so like any way that we organize our state in layer too, we can do thaton the layer one because it is general purpose, and that is the magic of youknow. Trin completeness and- and you know, machine simulation t sait's, ageneral framework of aggregating information, St Tust. Exactly that's that's exactly if we could just go back change thewords roll up and change. The Word Plaza to be something that had any anytie to what they actually mean in the fundamental properties of them. Butthen the eth maxis couldn't show it in three words: On titter, it's at old, mergl re anexactly, Oh boy, right we're going to be USINGTHEI.Recording that's in the in the h. The meeting here Ti should be fine,I'll just do with him postprossing after that Yho sounds good. You are recordingagain and we continue srark for that yeah. So Li Um, I what I worry about now now that a few of theseimplementations have actually hit painnet and people are using them uh the security involved. With thesethings like how can we be sure, Li e? How can a developer, who wants totry and utilize these leterto solutions...

...for scaling, be confident thatsomething isn't going to happen? It just breaks the whole thing, becausethis is really new like. Why should someone start to then develop usingthis technology that allows it they're into scale? That is a great question and, to behonest, there are, I don't, have a perfect answer. Um there are a numberof places where these protocols can break, so they can break in the kind offundamental specification right or they can break in the verification of thatspecification, ECA. The actual implementation does the implementationmatch the Spet Um now for the actual specification Um, the way that adeveloper can be sure I mean, of course they can read about the architecture. More than likely, they will rely on thesocial signaling of people who Um have been established. As being you know,experts about these kinds of architectures- and you know we'll kindof follow what other people say and that's totally fine assuming we doactually like you know, provide good, provide good. You know Um, like resources for people to look. Youknow, Oh, this person has evaluated this framework, and you know this. Thisconstruction actually makes sense and follows the you know and provides theseguarantees under these security assumptions and, of course, H, likesecurity is only um in relation to a threat model, and sopart of this process is going to be kind of figuring out. What is asensible threat model, and definitely even the experts really disagree aboutwhat a sensible threat model is for different constructions. So it's notgoing to be clear, cut to figure out what construction makes sense on a whole, you know roll ups, that youcan at least get the guarantee that you can exit from the roll up, um without h being censored. That is something thaty you would like, and by exit I just mean go from the roll up into layeron.That's like that's something that you definitely want from your construction, and I you know hopefully, hopefully thespectification makes it very obvious why that property would would um bepreserved, but then there's the implementation and that that's a wholeanother nightmare, because there are so many places where these protocols cango wrong. In fact, if there him itself, you know it was aninsane project and it's a miracle that it actually was put together and didn'tcontain more bugs than it already had, and so like this process, you'rebuilding like anetherium inside of etherium, and so of course, it's goingto be a very difficult prospect to get right and then not only this. What happens when these protocols gowrong, because how do you actually coordinate a migration on layer too,when it's much more difficult to fork? If you have layer, one assets locked upin the layr too, you need to build upgread ability into the layer, onecontracts to actually move over to the new chain. That has a broken feature,which of course it will. So this is like it's a hairy mess when you getinto the practicality of you know, implementing these protocols without huge amounts ofbugs and having a reasonable upgrade path. Where is that responsibility to lie? Isthat is that, like that, tecphnical debt have to be absorbed by thedevelopers? O are building these things, or is it or does it is it? Is it addingextra responsibility and understanding to the end user? That is a great question Um, sogenerally it has. It seems to be the case that end users will just kind oftoo soon extent blindly follow what developers put out and then they kindof like what projects get some kind of...

...social recognition as being relativelysafe and secure, and so in the end, I think that the onus to a l for a lar toa large extent is actually on the dapt developer, who is providing a who is mkind of using the layer too and h and and kind of maintaining if they aremaintaining or or suggesting? I guess is really the word suggesting aparticular bridgh between the layer, one and the layer too. So one thingthat we have been thinking about Um is essentially you know the the you know when you deposit intoa layer too one thing that you can do, that's actually kind of that's helpful.Is You can deposit instead of depositing the AC? The actual token youcan deposit a rapped version of your token or something like that, and in that way you can kind of get alittle bit of safety assuming you're rapped token itself has some upgrade.That is sensible. Of course, if you are like you know, using a kind of more pure approach- and you don't have somekind of you know- upgrade mechanism on your smart contract, which I a you knowprobably suggest, then you just have to be ready to, like quote hardfork yoursmart contract and like migrate over to some other chain and point your frontend to some other. You know balance set of balances if in the worst case, sothese are just like things that adapt developers are going to be. You knowneed to be aware of when theyare migrating two layor two protocols- it's really interesting, how much it itstill does kind of feel like the same thing all over again, butyou've moved it up a lar. The way that you, you know, you're talking aboutjust having social consensus about Oh crap, like this ebm inside the EBM,turns out to be fundamentally broken. Let's just all go to this other EBM INS,idbm um, but when I wanted to, I wanted to hear a bit more about whatit is that that you guys are doing. I think a good kind o like prewreck forfor this or like useful Um companion episode, is the one with John Avlerfrom before. I was here, but helped me get up to speed on rollups, and so my understanding is that they'redoing something a bit simpler, like their state transition, is basicallymaybe like year, ee twenty tokens only just like sending basically sendingreceiving Tokens Um, as we've alluded to you guys, are doinglike afull on Ebm h. What Os? What does that look like? How have you actuallygo about implementing that? And whereare some of the challenges there Vera question, so it is first, it is anightmare to implement these things B, but nonetheless we we persevere and wetry to find the simplest possible approach. So at a high level Um all ofthese roll ups can have different quotes. State transition functions, thestate transition function of Aljan is the evim state transition function, andso what we wanted to do was we wanted to preserve the developer tooling andthe you know: State transition function of the evim and so h we have beenworking on this thing called th Ovm, and this is you know, the a set of smart contracts. Interestingly,so so the ovm, the name right. It's really o evm to an extent because it isan evm that can be executed optimistically inside of an optimisticroll up. It technically t can also be executed using PLAZMA, but that's wayharder to Implebet, and so the way that it works is essentially wewant. We create a small like sandbox kind of smart contract sandbox that hasall of the different functionality that you'll see in the evm from createcreate two all the kinds of opcodes,...

...but notably, we do not actually createit's, not a machine level virtualization. Instead, this is anenvironment virtualization. So it's kind of the difference between you know.Um, you know vmware, which is machine levelvirtualization, where you kind of actually ex. You write all the opcodesin solidity, for instance, and you execute it there versus docker, whichinstead just crereates a kind of self contained environment that runs on yourbare metal, and so we we, we said: okay, it's hard enough to build anything. Wemight as well do the thing that requires the least you know, changesand which you know not. You know not build of evm inside of the EVM. Instead,we can just create this like virtual environment and so kind of on a on alow level. How that works is smart contracts, the the off chain kind of chain, the OP offtaneoptimistic roll up. You are able to interact with it almost as if it waslike an etherium side chain, so, like think of it as like, a t like a Rinkabetest nut right, it's not it's not the same provider as l one, but you knowyou can deploy contracts just the same. You can you know Commu, you can you know trade, yourbalances just the same, except it has one extra feature, and that is you know,l ontl, two communication, so you can actually send a message from L, L, oneto Ltwo and you can send a message from Ltuda l one, but that's essentiallywhat it's doing and we execute all of this off chain in a way that if it wereto ever go wrong, if there were ever to be an invalid state transition, thatwas, you know, executed off chain. We could go back on chain and prove thatparticular state transition. Te was invalid and the way we do that is iskind of similar to the technology. That's used in stateless clients, so wewe show the we deploy all the contracts that were touched in that transaction.Deploy. These storage slots, you know, prove the storge slots that weretouched in that transaction and then actually play that transaction on onlone, and then we just get one. You know one execution of a transactionthat we can say. Okay, does this match up with the you know, what was postedor is it? Is it fraudulent is what was posted fraudulent and this? This isbasically to to allow us to M 'cause of I. I didn't really go into thedifference between zk roll up and optimistic rollup, but optimistic roleup the way that the state transitions are. Are you know the validity of the state transitionsare preserved, is through these fraud proofs, and so we have a fraud proofthat executes a full. You know evm state transition function, Um and yeah, so that that's at a highlevel what it does and th the goal is to keep the developer tooling ofEtheorium, because the theory Um developer tooling is horrible, but it'sthe best out there Soso. We, we just need to approve it, but I' I'. I lovemy little evm contracts. What's the cost of doing that m thecost of leaving and proving an on chain where's IT scale with? Ah So it scales with the number ofthere's the number of contracts that are touched and the number of storagelots that are touched m. now, notably, we don't. They does not scale with thestate of the off chain system, and so, as long as there is, you know an upperbound on the you know, number of contracts that you can touch in thesize of those contracts, as well as the number of store slots in the size ofthose stora slots. Then you can establish an upper bound on the actualfraud proof itself Um. Now the the fraud proof will it if it is proved invalid. There isyou know, of course there is some cost to this, and so you definitely need abond to be posted when you actually...

...submit one of these STAE. You know manyof these state routes and you submit something invalid. There must be somekind of you know: Punishment to the party that submitted the invalid stateroute, and so that punishment, you know some portion of that H- is- is mactually sent to the person who proved the fraud and hopefully covers the costof the actual fraud proof. Is there an? Is there any worry aboutlike over economics, something like this, where the cost of proving a fraudis too high to c really do so, because no it's not worth it. Yes, and in factthat is one of the big reasons why Um, a a kind of small, optimistic roll upchain is a little bit more dangerous than a kind of big optimistic roll upchain. It's it's! It's very it's interesting in this way, so the Um, if you have a kind of chain that no onecares about, then you have fewer people checking you know, sinkingsynchronizing that chain, so there are fewer people who can actually detectfraud in the first place. You know if, if no one really cares to sink it, thenclearlyffrod will just get through without without thinking about it andthen the second thing is: If there is not that much value, you know to belost or if, if a disruption service is not, actually you know worth it toanyone, then Yeth, there's, there's a possibility that you know spending youknow a few hundred dollars on on a frokproof or more is is not even goingto be worth it, and so this is like there is actually some economies ofscale for these for these chains for sure the Tes, the Trim God Johns heartwell yeah. This is probably t a stupid idea, but have you considered likehaving the bond paid in gas token so that it like flows with the the network congestion? Exactly exactlyso? We definitely have. In fact there are. We need to generally rate limitthe transaction submission to these chains, and so one interesting way torate limit is actually you know charge some kind of you know some kind of fee.In some sense you know, like S, Burns burn some amount of of value, Um kindand that burn could just be like buying gas token that is distributed to theperson who Um proves the fraud. So so so you can even like establish a kindof like reserve of all of the gas token that is like m held and then th. Thebond can just be in you know: Gast Oken and pay pay the person back. So yes,this is. This is a great it's honestly hilarious, but it's a little funky, butit's pretty cool. We're againinto, like I feel like thatslike naval gazing mechanism, design, go o Acory, I shounto say like a you said:There are ot, somewhat kind of like TEC Economis a scale sociate wit Thi,trying to pprognosticate of the future of what the state of the THEORMICAsystem looks like and how you almost have like Emergen, emergent communities acrossdifferent level, layer, toos, and why they would be on the same one 'cause.It seems to e be worthwhile to have communities of people working on a specifically or totogether, because they have a kind of a unified goal and it it grows a lot oflike kind. The REASONI IDUS said you want there to be more people and morevalues. Tha, it's it's! There are enough people watchingthere's an a failure. Associat there, there's commerce alongside and then like. Why doesn't it just allco Lessento, a single lar too? Why woul W wh what s? What's the purpose ofhaving multiple of them? If you need economies of Scaleso like this,...

I think that there is incentive forther to be multiple layer, tos Um, there's like Um, you know, Oh, whosetech are you USAING right, there's like mthere's, there's many block chains.For the same reason, however, I think it will result in some kind of powerlaw distribution where you do end up, having kind of dominant layer to chainsthat most of the value floods to Um, and notably this is not because of akind of fundamental necessity for value to be to kind of Coless on the samequote: State machine or same chain. It's actually more an issue becausedeveloper tooling and like asyncheniss, communication and kind of like networkof blockchains is just so much more technically difficult to achieve in thenext year, two years, three years, maybe that it d, it's not likepractical for a developer to to kind of build composable applications acrossmany different. You know block chain environments, and so this is like sowhen I say it will coalescen to like one chain. There is a possible futurewhere things kind of spread out a little bit more because asynchronitcommunication and smart contracts and the kind of programming abstractionsthat we work at do more. You know intelligent load, balancing acrossblockchains with similar security constraints, and so that's like possible, but it is hard, and so everything is just it'smore about practicality. It's not about a feasibility, so yeah. What? What does that look likedo? Do you think that you might have like two or three roll up contracts andyou get like Defi on one of them and Games onanother? I dows on another like the ones that want to be colocatthey'relike the ones that want to be colocated in the same way that all the shoestores like to be in the same area in New York or something and then so. That's that's one, and then too islike how? How do is there some form of Y, you mentione acingchee communicationthat that you can, you can bake up to to let them talk to each other betweendifferent? U Rolvs, yeah! I I. I think that it's not crazyto think that there will be like like services Um in in the same environment.That definitely seems reasonable. There might be multiple defy you know, chainsor many Ma, many forks of the same DVI products across multiple chains andlike it's it's more about like whereis, the value right now Um, and so that'slike Tho. Those are all possibilities. There are definitely Um. It is definitely possible to doasynchrenist communication across Um roll ups and in fact you can have you can even build kind of roll upchains that Um facilitate that asynchronistcommunication a little bit more, but they they okay, so so zk roll ups. It's actuallya little bit easier to do. asyncherness, communication across chains, becausethat asyncheniss communication, you know that it is valid the moment thatit is posted so like when you post a new new. You know Zk Roll opstate rootand I didn't explain this IAPOLOGIZE thes optimistic and C K. Rola Zica rollup your proving upfront all of the statetransitions, instead of relying onthis frodproof now the frog prove is great because it gives us you know theEVM, but it's bad because it r requires a challenge period and that challengeperiod is like a week. So if you want to send a message from optimistic rolup to layer, one in asynchronos message, you can do it, but you have to wait aweek or I mean you can technically wait...

...less time, but the less time you wait,they'll less secure your message, and so I just you know, suggest a week andthat's based on kind of how much therium has been doss before it's beendoss for three days and then you know you multiply that by two and add a dayand that's you know, that's your asnconess communication. That's yourtime out and that's from verry white hat. By the way. That's th t that's agreat great little one, but Zk roll up m. You prove the state transition upfront, which means that a message from one's EC rollof change, O an other s,Ecaol Chin can theoretically happen. You know the moment it's they're postedto layer, one Um. Now the the that means that that does imply that ZC thatthe second sec role of chain kind of intraspect, some information about thefirst one that you're sending the message to kind of like crosslinks inetoo, and in fact these are M- you know very similar concepts, and so just likethere are going to be. You know many there's many shards, there's also youknow many roll ups and these roll ups can communicate through crosslinks andthe moment they are finalized. The communication can go through Um and they're, in fact, they're they're,very similar in how how t how you reason about them. So, yes, asinkerness communication isvery possible, Um yeah, and so just just you said, likeyou know, if you have optimistic rollups, you have to wait for the the fraud proof. The PR the provingtime that makes sense is that is that, like delay, hardcoded, so that you knowlike my role up in in contract x BC has to be like no like different roll upsin where it might be coming from and be like. Oh, we need a way a week for thatOneso likethat. In fact, in fact, it is nothard coded, it is solely up to the layer, one contractthat is kind of executing the withdrawal to determinethe delay period of the you know underlying roll up. Now you can have itso that h there are, there are prob. There areways to to kind of Um not to to make it not H. configurable, like you, candesign a role up that kindof like locks, these things down and so they're,probably their ways to like mess around with is but like in a well designedroll you, the layer, one contracts can determine so if they think, if you'refor your application, you're like Oh, I only want taway to day. You know youcan you can just wait a day in fact hilariously. You can also do weirdthings like. Oh, I want to trust everything that comes from this roll up.I can trust it immediately. If I get this, you know get five signatures from the fivepeople. I trust then I'll. You know mint the asset immediately and considerit finalized, Um and, and that can be a totally that doesn't have to do withthe roll up. It's just like a totally separate. You know multisay that peopleare that people are signing off on or you know you can you can you knowtokenize these these uh, you know kind of instant exits it'. It gets crazy, um right right, so the delays in theoriginal contract itself, exactly I dos Ese know these things going intogoing into givean application, like you said, like they kind of just, have totrust that whatever application they're using Um made the right decisions based on theirsecurity or at least earleast, informs them on all these things, but like baseon what you just said, there's a tremendous broad range of options that mayeventually be applied to all these different Rolla instantiations Ho. How are you going to advocate this? That is a great question M and, in fact,M my Methoo, my like mental, my my approach, my personal approach, andthis. Even when I talk to people, I...

...very much respect like I, I they mayhave different opinions, but I am all four like maximumly giving power todevelopers like. I don't care that I I know that like in in practice, thesekinds of things are going to become unbelievably confusing, potentiallythere's like a potential for it to be just absolutely horribly, confusing youdeposit, eath, into a layor two that has a withdrawal period of five daysand then you deposit, Eth, into a deposit contract that is which RallEriod of one day and now on that chain. You can't think that those eth thosetwo different types of eath are fungible 'cause one has a differentsecurity assumption than the other, and so it's kind of insane Um, but I kindadig it and I think that the the kind of real person to answer this question isgoing to be. Like the you know, amazing UX designers, like you, know, Colelfrom uniswap, or something that that I I comes up with a way to organize thisinformation in a digestiable. You know manner and likethis is not just a problem. That is like new to the space by any meansright like the fact that we sign these meadowmastransactions and often timesyou know. In the early days, we had no idea what we were signing like. Noteven it wouldn't even tell us, it was an ears twenty transfer, it was justlike. Oh Yeah, I'm signing a you know ABC one two three and it's ridiculous and you're just totallytrusting the front t and so like. I think that one thing that people mayneed to get more accustomed to is that using adapt is not just about trustingtheir front. I mean notbjust about trusting their more contracts. It'sabout trusting their front end and front ends are really really in control.Um. I am I I would say that I feel pretty confident in my ability to likeread. You know transaction data, raw transaction data, but I am veryconfident also that I could be fooled into signing a bad transaction thatdoes not express my intent, and so I don't think that there's really any wayto get around it. Now, like a question of like block explorers, like you know,ether scan is going to have to figure out. How do we display this informationin a consumable way and dapts themselves? How do we explain thisinformation but to an extent we have some early examples. DIDX is an earlyexample of a project that you go to their their their website. You depositinto some thing that you don't really understand, and then you start signingmedotransactions that you know affect the state of a of a of a you know,state machine that is, you know, kind of a little bit less standard, andso it may look something along those lines, and I I faith I'll agree with you at least that Um optionality for developers is good. That way like, hopefully, you mean in the end, youhope that good good developers understand their users and then andthen accommodate the available Tich to their users and the more options theyhave. The betterey're able to do that so, like optionality, it ust gives themthe ultimate Um availability to cater to whateverusecase they're trying to solve not having that forces them to kind ofpigeon hold themselves into what's available, and I think that's kind ofLi e hallmark of a theory in the first place is, you know, build buildappropriately to what you're trying to do chip. That's a lot of stuff, I'm tryingto thinkit as a company like working for a company that tries to solve thisproblem. Providing an interface to atheorium my mind is: is racing in terms of thecomplexity complexitis rinvolved, of how this aunt landscape is evolving, aswell as like associating risk and then relaying that to the user. Yeah terrifying t these early, theseearly experiments with layer, two are...

...going to be very weird, veryinteresting. Like hacks, you know broken front end, just total confusion about what is what,on the side of you, know the users it's and developers. To be honest, it's it'sgoing to be a nightmare, but it's going to be a fun one and we we definitelyneed it. We need to go through those growing pains, Um, there's only one wayto learn. Do you? What are you worried about stared at me this ole time, if hisearid through thm everything I mean, do they like? I'm, like I like the idea of layer tobut every complexity which sharding bringslator two makes even worse, because it's like sharting at least everythingis essentially part of one network. You kind o know where and how to find stuffwith with all these lair two things. How do I find all this shit like? How does just how does myapplication like if I build a fronten for this? How does my application knowwhat notes to connect to it Goin to be horribly annoying, and Idon't see anyone working on standards towards ike kind o do that, so I thinkit's kind of important to establish those as well doesn't Yeans to solveall this tin I mean, ideally, I don't you could probably solve it with you.An US right like you, could buy the domain for your network and then incodethat, but then it's also like authoritative, like whoever owns the contract owns network.It just brings all these other complexities beyond that yeah, but in the othersense, like the whole concept of like, I don't needa permanent record for my dognut purses like later twos. Give you give youcontext to operate within a smaller subset of people than the global state and Jane en you're.Trusting you're ruaning yourself in a larger trust, which I think is the madegoal here. That is fair, but that smaller subsetof people needs to find that context, yeah an that's UPTA at opers inmarketing and so on and so forth to be able to advertise that provideit appropriately like it's it. This is, I think, just the waythe world works like you're going to have some sets of people, don't want todo a specific thing and they don't need access to the inefficiencies of like globalconsensus in order to do that, but they would like some type of trustless rootso that they can operate with. I guess it roperate in a more trustlessenvironment than what they traditionally have to now and an the ability to kind of haveliquidity across the entire network, while operating in these smallersubsets of people. I mean think about the situation of like the I don't know.That's almost like what I used to argue with whan plasma was becoming a thingand how I kind of imagine this entirr space growing, and that is the SE analogy of the way the Internetgrew ul, respector corporations. It was mostly just the lands right, you had.You had a land and then companies were very sceptical to join the Internet becauseit was a gross CV place for the longest time, so they do they uilt their ownInternet. Just an internal corporate corporate in Ritd to land evecuallyfound a way to establish a secure link to the global Internet, so they canactually talk to each other as corporations, and then the Internetjust grew and grew and grow and grew because, like you said, standards grewyou have better staky, protocoll, so on and so forth, and the the bestpractices of being a proper citizen of the Internet got alittle better. But it's not like lambs went away. Everyone has a land. Yourrouto provides you of Lanb in your...

...house, but you still have these things,but you just understand where the gaykeepers need tobe and layer toos are exactly what that are, but that Wat that is, in myopinion, therethey're the the lands of today's Internet, where the openpermissionless networks like a theorym or the Internet, and you just need tounderstand where the gatekeepers are an what the context is for how value andcommunication flows through them a I don't. I don't see that changing at all.It's just now we're dealing with value. Is there indivision to believe anythat's wrong? Carl? I really loved it. I really loved it mthe entire concept around like standards and coming up with like whatis what does security really mean and establishing that right we have Um. We have a a momentous amount of work todo to establish what a secure block chain really means and like what we canaccept from a secure blocchcain, and how do we establish those secure m kindof portals between these different different domains? In fact, by the wayH I was I was asking asking vatalic about like what's a word thatencompases a Shard and a roll up, an a Plazma, and you know all thesedifferent kind of state spaces, and you know good what he suggested was stillmain and I feel, like that's like very, very reasonable, like these are allkind of different, different domains. In the same, you know in the same blockchain in etherium that should all have pretty similar security m constraints,and the moment that you break some some, you know security assumptions we justhave to like prune those puine those networks and just basically say. Okay,you know we only accept a certain, you know a certain standard, Um and- and Ido think that we will eventually figure that out and it's not going to be it'sgoing to be a messy process, just just kind of as you describe cory like not it's not one person sayingthis is the Internet. It's just a kind of emergent. You know, encompassing ofmany domains andsa today a lot of it's going to begross yeah. Unfortunately, there is too muchponsi activity on detherium for anyone to be h entirely comfortable with, in myopinion, is there still that many pansies otherthan Hfag dfiis a good example? But the there aremany Eah. If you look at H, osgaxes there Um, but HEX Aysn, not a theory.IIT yeahnearses, NY MHM, Don Tay Laughe there like Bitcoin, fork thing n, it'sjust it's! It's a proof of inownership at a certain date that mince more heck.That goes, though, Richardard yeah, so pozsis shall not be named. Do NotInvest in panses Um. So one thing that, like I hear when wetalk about all these different Laer tos and Um the delay levels of abstractionbuilding on Etheorium, I start to to think that perhaps it's getting to betoo togh heavy like like how much can we secure how how many tokens can we secure withone underlying token? I think we just you know recently have started totransact more like, like recently the volume ofUm you see. Twenty transactions is assurpassed eath transactions, oreath issuance, so probably issuing so you kn. It gets tobe like if we really really get past that it doesn't. Take you setting up a lotof transactions to make it easily profitable to h to do some double spens and some somet.

You Know Hash Hower, mipulation attackslike like and so especiall. I guess like theresthere's like a phrasing that has been used, I've seen use with Um roll ups, which is like you know youget the same, guarantee security guarantees like you inherit thesecurity guarantees e Blair one and your layer too, and l ke well. Can youjust copy and paste th the security of this proof of work, Chan and double itlike that sounds like pretsecurity to me, Umand, so I'm juriously might haveto say about that. That is such a good point. Oh my goodness. The littletalked about problem with all of this is the parasitic l too problem, because Um later one is, of course youknow secured by Mainchain otheorium right. You know all all these minorsthey make a bunch of money. Now one thing that scales with the value on the network is myv or MINUminor extractable value, the more value on a network, the more money you haveto gain to you know when you arbitrage different, you know coins or you know,arbitrise uniswop against something else or you know, there's just freemoney on the network, and we see this all the time with front runners. Frontrunners are like some of the top gas guzzlers on etherium Um. Now this M ev, when we start buildingthese layer, two protoocalls, actually moves into later too, because the youknow, the value that you can extract from front running and from like youknow, being first in line, is actually going to be extracted by the layer, twominors, not the layer, one minors which okay, yes, the lar twos are committinga bunch of data to layer. One posting this for these transactions etcetera,but a massive amount of their h. You know of the layer. One value comes fromthe fact that a minor has the unilateral opportunity to front run um now. What does this mean for layer onewhen nothing is going on of interest in layer, one outside of just being?Basically a kind of big data? Dump of you know, transactions, and, especiallyin eath too, where you end up scaling up how much data you can dump and that,of course, because of the supply increase will reduce the price Um. Sowe have this weird prob problem where money is going to be extracted at theedges. The M V is extracted at the edges of the network and the inside ofthe network is going to grow and reduce the price of you know the those likeyou know, roll ups posting tr h. You know transactions to layer want. So ifeverything is using, if we're all using roll ups, which it seems very possiblethat you know if therm becomes just, you know a hub for many differentrollups and plasmas and whatnot. How are we going to be confident that theelwon minors are actually making enough money Um? So this is this is, like youknow, a crazy, you know question and who knows? I think that there there area you know. Ideally you know proof of steak. You know there's this likesocial layer that kind of kicks in that you know it's like a too big to failargument. That's one thing that I c honestly like even you know for forwhat it's worth. I don't know if too big to fail has great connotations, butnone the less like if the main chain oftherium has some kind of big disputeover what is valid like the social layer figuring it out. Is You know, Hit's better to have one big failure that everyone like we need to figureout tomorrow or everything's Gong to stop verses like intermittent failuresacross you know. You know, Oh last month, this guy failed, Oh this monththat guy felt, like you know...

...the overall reliability. You know I would. I would be more concernedabout so these are. It is an open question, though, and my arguments foryou know why the parasitic ltwo problem is, you know a problem yeah. I I I can't convince you thatit's not going to be a problem, because it's it's a little. It's a littleconcerning now. One thing you can do, and one thingthat we are are thinking about and planning on doing. Um is actuallytrying to design mechanisms to extract this Mv from the layer, two minors. Sothis is. This is a kind of talike, a tangent to to layer too, but it givesus a chance to redesign mining incentives, and so one of these ways toredesign it is is introduce something called Miva, minor, extractable valueauctions and the long story short it's his interesting mechanism forauctioning off this Mev, and so one possibility, and by the way, the reasonwhy you want to auction off this en M V is 'cause. You take it away from thelayer, two minors who are you know questionably securing the network likewho knows, if that's really, it's really worth it to pay them that muchmoney. You now have a pot of money that youcan use for something and that part of money could is, is aperfect the way to use it really, the only reasonable way to use it is forfunding public goods and solving tragedy to the Commons and right now,security is massively overfunded. I think that that's like pretty clear in the future, where we moved to eathtoo and a bunch of roll ups. It might be underfunded, but then, hopefully wehave like social coordination similar to you know, t e, the socialcorinations of taxes and government that kind of push money back into thecenter layer and like provide money to secure the network. Now it's a little crazy of a future,but that's just one picture of it like to push back a little bit. If, if that value flows to the edges as bepotentially move into a ecosystem where most people are operating on layer twos,because it's more economically feasible for them to do so- and the mvt likemoves in that direction and and in the associated value in prices of lar, onedrop significantly. That's a market, the functionality of layer, one isn'tgoing anywhere and eventually those uscasses come back to laier one and so like. There's no reason tobelieve that people won't just use layer one whenever it's the costly l, it's beneficial of themto do so. Maybe it's a issue with kind of how anapplication is built if it has that functionality or if you want to playthat market on intergetting access for Spurs in certain types of use cases,but once again, like that's, not necessarily a bad thing, because oafter all, like the underlying, like fundamental value, is in digitalscarcity and when you introduce layer, twos ou know this type of thing. It'snot like you'R l, the digital scarcity isn't growing or shrinking differently. It's just being moved anddistributed across different things, with different constraints on how itgets moved and so like there's AV still value in the scarcity, because thescarcily doesn't really change. It's just how it gets used, as is there'smore options and more Coni gess kind of constrains. And now it's move so like.I think, that's not too big of an issue if things get cheaper over a certainperiod of time, because the functionalityes in going away, it'sonly getting expanded. People are going to find to way to use it n and if, ifit solves the front running and and backrunning, which is a argualy moredifficult issue, Um than good because like the baselayer needs to beefficient, and if we can move some of that kind of behavior from market players to docertain things in certain types of situations to an off flayer situation, same t situation, teaen times, that's agood thing and we can deal with the consementis there. So I'm all for thatcaus. Once again, Lik functionality...

...isn't going away. layor one's stillgoing to be able to do all the things, but then you have extra options todoing elsewhere if it's too expensive to do an other one, and one one like interesting note is that, in fact,this problem is not a hundred percent. The problem of Mev reduction isactually not one hundred percent Um tided to layer tos in. In other words,you can actually reduce MV significantly using smart contractdesigns on layer ones, Sig just only on laer one, and probably the mostsignificant is. If we get to a future where most transactions are shieldedblinded, you know you, minors, don't know what the contents of thosetransactions are. Then it gets to the point where you you know I it's notreally as valuable to be a minor in those cases, and so, even and by theway, even in the kind of MeV auctions you know, modeled, there are ways totake away the MV from the auction, because ther're just generally ways toreduce MEV and so yeah. I it will be aninteresting, interesting future and I totally agree with your point that,like people will go where that Y, where the transactions are cheap hundred percent. Well, I think that what you justbrought up as an entirely new episode, so I ht it might be best to to to start torap up here or E. is there any question that you wish? We would have asked thatwe didn't hm. I guess the probably the biggest thing that I wantto learn, and maybe this is not a question. This is a question for your,for your audience. Toi extent is like Um. Actually, I honestly, I don't have aquestion. I tried really hard. I tried really really Ahere it just didn't come.I thought you asked great questions Um. You know, and I had a great time so I'mvery grateful for e for the whole thing. So then appreciate that where do peoplego to find out more about opthiism and and yourself umso optimism? Dot iois ar very youknow, shotty little website that we that we don't give enough love to Umand our docks are also similarly shoddy, but our get up repository, it's veryactive. We try really hard on that guy. So you know optimis optimism on Gidub,also on twitter m optimism, TV C and- and I am Karl Dattec- you can go to mywebsite I'll post, more blockposts when I finished this darn project and we youknow, improve the scaleability on Atheorium, so thalot EETON. Thank you awesome. Thanks for.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (109)