Hashing It Out
Hashing It Out

Episode 1 · 3 years ago

Hashing It Out #1: Nick Johnson

ABOUT THIS EPISODE

Part of the Bitcoin Podcast Network, we have a new show where we interview Nick Johnson of the Ethereum Foundation. He talks about the Ethereum Name Service (ENS). ENS is bringing human-understandable naming to address resources both on-chain and off-chain. Nick is incorporating a new non-profit organization to support the efforts to build out a decentralized system for creating and managing these names. In addition, we pull from his expertise as go-ethereum engineer to get his take on dapp usability, solidity language quirks, blockchain scaling, and our blockchain-driven future.

Now, injurykindwork welcome to hashing it out APOCASP forRetolk to the tech, innovators, behind blocked in intrastructure anddecentralized networks. We dive into the weevs to get at Wyan how peoplefilled this technology, the problems they face along the way come Listeninand learn from the best in the business. You can join their ranks a guys, welcome to episodes one ofhashing it out Um with your hosts, caler Qarry, petty and Callan Cuche,calling you and say hello, ello guys, hostes numbers, zero and OS number one sure Ra spartment zero,so hash get out more along the lines of a podcast that wants to dive intotechnical aspects of block Tan technology, T sitituleter technology,whatever you want to call it and for our first podcasts we wet moreappropriate. Did you bring on Nick Johnson from a theorium greader of theetherium name system? Amongst many other things? Nik, you want to giveyourself a quick introduction. I YP as aluded I' Johnson, an my mainproject is the Athereum name service which I created about eighteen monthsago. Sorry about a year ago, we launched and it serves to bring usefulhuman and understandable names to the block chain, specifically etherium.I've also worked on goitherium and other projects like community out reachand so forth, very cool yeah. So actually W kindocurious about that. I haven't followed the NS project much since about June,maybe early July, and w what kind of happened with that. Maybe Yo could tellthe story. I know that you were on the episode one twenty nine o the doint podcast. It was a great episid byway I relistened to it again today I was likeoh o yeah, there ware so many things that actually just came sure in that IVsaid. But one thing that I haven't heard a whole lot about is basicallywhat continued on with the h the NS system, and I saw that the lastbasically state of the syster was in June mid June. I want to say so: What'swhat's the current state of the Ns and how are things going to what's e INSnow hat's? What's going on with all that? Yes, so we we had a a reallysuccessful launch, as you alluded to we sold about or er rented out about ahundred and seventy thousand dmanes. Since then, we've been sort of headsdown working on more features so with deployed a manager EP, which is youknow much like you might be familiar with to manage your Denis names and andrecords, and so forth. Bat Freinis and we recently deployed atil four and veryquickly, allocating yourself ta subtemane, so namal Wallet and Wenclicksort of thing, because you know usability- is one of the big issues wittheyre, focusing on at the moment. We've also worked with a lot of clients and otherother debts and so forth to get Bini support added so got over a dozen H,different eplications supporting UNIS now. But the really big news is that,just very recently, an infect on going now we're splitting off from the TherinFoundation into our ownno profit organization, we're getting independentfunding, starting with a very generous grant from the Theim Foundation andwere able to for the first time hire a team. So it won't just be me withvolunteers. ITILL be a full time team working on andproving in it, and sowe've got some big goals there. Oh that's also what kind of improvementeAr You'e looking for so sort of top of our list is we we launched workinggroups a while ago to sort of get the community involved in IMPREVN IIAT,specifically coming out with version to the NIS Registra, so the permanent waythat people will rejister names on the system and also, if it's like,integrating Enis with Deani, so that you can claim your regulardny tomaininside etherium and other other efforts on those lines and we're also in termsof doing our own development going to completely rewrite the Enis manager andall the existing tools to integrate them into one sort of cahering package.We've got a UX and antderection expertin who'se helping us together acheringtand, and you know usable AP ecause. Currently, things are butfragmented and basically we're focusing really heavily on user experience. Wewant to make it as as easy and straightward as possible for people toget started. Fhan people to use it. Our medium term goal is that noetheoriumuser should have to enter a block chain address unless they're developer,indusers should just be able to deal with names the whole time, and thatshould be all they have to know about yeah. So what were the lessons youlearn from the one point of version, specifically so the biggest one is. Weknew that the aupction proces would be.

You know complex, and there would be.You know that that some users would find it difficult, but weunderappreciated how much difficulty it would create. So you know, although the PRAPESES insuracs of their pricefornames, it's also, you know it's easy to trip up and make a mistake with it.So, though, those are the sort of lessons we thought, we'd learn andthat's why we deployed an inter run rather than launching Straihtin andsaying yes, this is the permanent set up, but we're trying to you know takethat ND and use that to build a BIS, long term system and and that'sprobably going to involve making it possible for people to register nameswith just a single operation rather than having to go through the holeloion process. I liked Athe Tok a few steps back and try and figure out orlike explain to the listeners. Why is this system necessary and what makes itbetter than the DNS like? Why can't you just use the traditional services to dothe same thing you're doing now? What makes e Ns better, then then othersolutions and and how far do you see that going? Are we still going to needtraditional solutions, or is this a complete replacement MHM? So there's a couple of main advantages andat sort of depends on your needs, but first of all existing solutions likeDenius, can't be applied directly to to naming atheoreum resources. You can dodenious resolution from inside smack contracts, the ram appropriate recordtypes, nd and standards for doing that sort of thing, but, more importantly,doing it through smart contracts d and on the theorium chain, gives you theability to to enforce restructions and so forth, so that you can makeassertions like. I know that I am the only person with control over Thi thisrecord. You know I I almost damaying in the truest sense, but I would say that the the main advantages is building a a naming,systemthat sot of native toetherium, which means that any ethereum clientalready has the resources it needs in order to be oble to resolve these nameswithout having to reach Yau to external systems and with you know, the completesecurity of the blockchain. So you know the records you feature accurate andbecause of the same guarantees that allow you to know that yo w your fronsare being double spent, and although you can achieve some of that withgenius and Deanis Sik Er few people actually use the INISCENIT's poorlydeployed and paulianforced across the entenetes. Although one of our otherefforts is that, if you do just want, you know in existing Denan that we'regoing to make it possible to import that into inus, so we kind of look atit as offering the thease to both worlds. If you want robust censorshipresistance, then you can use your dotith demain that you waught from theauction registrate. If you just want convenient, naming with your existingdomains, you'll be able to import them and and use them there, and an effectwas seen some interests from Dnes registrals who want to use the UNIStechnology to beak up their digines Tomans so that they can importall therecords the roote own and to atherium and then serveit from anywhere in theworld. Without me s of Dioisatex, taking it Theen and so forth, that's extlly quite interesting andthat s that's more than just like a redirect service. Correct, yes, Yeh! Sothe idea is, you know normally witha typical denious registry. You have theroot zone stored on some set of you know globally distributed name servants,but if theyare taken down, then you kN W or made an accessible to you. Urauteof luck with the root zone on the theory in Bocchang Eone, who is anatherium client, a not that saxs can access the complete rood, owne and andresolving record without needing eesttose service, ythat's very cool. So what is what isyour timeline on a lot of this? Why are you going? How do you see this kind ofplang out in the near future to the far future, so we'rewe're, going throughthe whole incorporation process to get set up and fundeing everything rightnow and hopefully have that done and theyx weak or so, and you know, makesome big announcements about it, and we've already got most of the staff weneed V signed up and agreed to work on US and commited their time to it. So we're ptly hoping to get started onthis this week. Basically, in this coming week, in terms of of producingresults, an I guess, the outiline, for the main, the add of the new P, isprobably on the order conservatively of about six months and just because wewant to get thus right. We we could rush straight into riting, the air thatwe want to make sure the user experience is right, mature, it's it's!You know, hetting on firing on all cylinders, Musability wise in terms ofsome of the other efots like the denos integration and the the permanentregistrar. I think we can expect to see you know tat, seeing results a lotquicker than that the denus Registrat for instance. So you can register yourYo know your dot com or its Lazy Toman in thes is mostly there already. Itneeds some some polishing and some you...

...know small improvements, Andin some Uifor it inteds register our Gol was always to deploy that circle two yearsafter launch so ot a year from now, but you know W, we intend to have a workingversion and fardemo purposes and so that people can see whatyou know whatto expect a lot. Sooner than there, so I might h e not have understood thatcorrectly. The first time I heard it, you actually want people to take theirdotcom, donames and use it in the the so h. How are they goin O actuallyverify? How are you, how is this system Goingto know, then that you knowCalincuce dotcom is actually owned by this wallet ID and that I you know and Ho. What's? Is there an ky se youassociated with this, like? How are you actually going to manage that and isthere some centralization compona that you're going to need to add to this inorder to make that habit? So it's the magic of Dinis Sik and we we can createa degining sick oracle contract on the chain and we've already written one.That knows the rout keys for Deanis, the ones that are controlled by youknow, Iana and so forth, and, given that you can you V use achain of trust to prove the existence of any Dinis record that signedcorrectly using theinissic as long as it uses one of the algorithms oftherincansupport, which is about eight hundred and ninety percent of allDunisit signd demains at the moment, so it introduces a reliance on the routekeys. You know the at fruit keys and it also introduces a reliance on yourRegistran, but that only extends as far as that section of Ns, if you like, sowe're, basically saying that if you trust the DNIS Registra to accuratelyreport your unituor to sign only things that are owned by you, then you can trust that your dimn isaccurately represented an ans and if you don't, then the the Doti RegistrarI still available for Yumen a trustless fashion. So we kind of vew offeringuses both trade offs er like Tem Choese, really cool, really cool. So this is anindependent organization. You just started you'R just incorporated. Isthis an open source project? Still is this something that you're looking forhelp with is is something that people can get involved on this particularproject Yep. Absolutely it is an open source project and the organizationwill be a non Profiba wfided by grants and will use that money too both tohire people to work ontiss when they can to hand our bounties and just togenerally sort of help run the community, but it's still very much anogen source project and we we welcome volunteers and so forth. The best placeto start with that would be discuss, dot. Eni, stop domains, that'our ourdiscussion board for the working groups. You know we really welcome and putthere in terms of helping design. The next componens ofthenece, RN, cool and you're. Looking y you're, really focusing on kind of theuser interface sided things, but what kind of technical like aspects of thisare you feeling like you were going to need to innovate on the most w a? Arethere any ousstany questions that are currently in the system that you coulduse? Particular experts on that might be beneficial to the project,absolutely. Ah, the main. The main stuff is focusing around the thepermanent Registra or Dotet alsort of the next yen registrat. So for thefirst version we had this option process and when you won an option,your funds were locked away until you released the name. If you give thedemain back, you geve your money back and that work well enough or forfairway to allocate tomains initially toug it sourd ave become apparent thatnot everyone has the same cost of capital, so some people have much. You know smaller cost tothem of locking away sonether for a year or two years or ten years thanothers and tha perversely encourages squatting. When we would, you know tosome degree when we were trying to discourage it with hat michanism, so th,the big open questions are around d. What should we do? Instead of that inthe most likely solution is to charge a a fixd rent pour domain rather than youknow, a deposit? How should we auction off new names when we first releasethem so when we released names shorter than seven characters? You know what hesystem should we use and if we're going to charge ren, how should we set? Whatthat rent is because it's pretty andelegant to have a disentrized systemand then have you know one person who cansic cad you laturally see I how muchit costs fo a year to on you domain m much better. If we can instead have thesystem self regulate so that, if you know there are lots of useless demainsbeing registered not used, for instance, then the prprice goes up, and ifthere's you know fewer domains that are you kN W or the use Te drate is veryhigh, then the price goes down, for instance, so where we could really useinputters from people who have you know clever ideas, ind building selfregulating systems. You know good game theoretc founded conceept cool Goo, and so I I guess that theKino leads to another question. That's kind of been on my mind. A lot latelyis Um, so you're you're deploying this on the main theory of network. How doesthe current price of ether affect all...

...this? What do you feel like it en youeven care? Now, that's cover you bring up the kind of the game. Theory side ofthings is the incentivation isaion model on this. If it's tied to etherspecifically an and yet the world is still in a Fiat world, how do you feelthat the technical challenges kindof justify the work this being done here?How do you feel like you, you kind of like like yeah, how? How do you feel like,like this whole scale, ability issue too like if you have a ton of databeing dumped into this one contract like if this hosting every singledonane and this would actually increase te size of the batching quite a bit? Sowhat kind of you know solutions? Are You seying that would ah not just about matter price, which isthe scale bi transaction fees, all this kind, OS UFF? How does that play intoyour system and what kind of concerns do you have rith regard to that? Yeh,so for the first question in terms of of flucteratting price, that's anotherreason why we really want a self regulating system because, as youobserve, even if the the demand on the system for names doesn't change, theprice of of ether can change dramatically and if we were aingtomanually sit up, then not only trusting some third party to set theprices fairly they're also they have to keep reaching and an adjusting as asthe price changes, so a self regulating system would steal around. Then itwouldn't care about. You know the price of Ethu. It would care about, say whatwas intage. Ofusers are deleging their domains and we would say: Well, we wantyou know to scen mains to be deleted ayear as useless or something like that in terms of the scaleability thing t along term, it's differently something we need to work on, because long term,an theor in itself is Goin to you know, is ginnadeall the scalability and UNISneeds to integrate with you know. If we have multiple shots, for instance, youneed to be able to resolve across shards and so forth short term. I'm I'mnot overly concerned because the volume of update data for names is relativelylow n. He the actual amount of DNISTARTA. Youknow I wouldn't want to store, perhaps the entirety of the Denese global routin in Etherian right mierbut. Certainly,you can store a Sizal fection of that Inne, surprisingly, small than out ofspace tay, you know named aterirs is quite small for your typical theoryumrecord, we're talking about maybe like one hundred and twenty eight bites per.You know per de main, so it takes an awful hot of domainds toadd up to a substantial amount of overhead, and but in the long run youknow as O say we're going to be dealing with the shouted world R or at least ascalable one, and it would be useful to be able towork with that and to some degree the ENI structure facilitates that so everydomain has to be recorded in the central registry contract, but only thewas the effect in its existence and the address of the Resolver, and you canwrite a resolt en that supports, for instance, state channels, ond frequentupdates, or you know, other scalability mechanisms without having to change thethe COR components Goo. So your actually are you more excited T. let'stalk about scale ability for a second TSAT, you brought it up and it's one ofthe topics. I really like m what what is wh? What is your kind of thought onthat? There's, a lot of work being done in very different way is sharting, asyou've mentioned Erewas. Also ineroperabilities, you get your pokadotin your cosmos and the Internet, a block chains, kind of thing going onthere. It was also your plagsman, your side chains, and then you got yourstate channels. What of these technologies? You really focus on mostand how do you see them kind of impacting what you're currently doing? So, I guess, from my point of view,shouting inside chains are sort of two different sides of the same coin. Youknow thee ones a homogeneous. You know multiple mirbogeneous chains, that'sshouting and the other is multiple hiterogeneous chains that sidechains.So I think the take for those is probably Gini mature at a similar rate,and although I kind of think that side chains are uneasier sell, because if you have a bunch of different APPS,it's easier to say, okay, we're going to tune. You know this chain just forthis AIP and this will be the the CIDRAKITI chain and then you know, CIP, pointed back tothe main of theorium chain and provide some way to swap resources and so on,thats a a lot easier than a world in which, like the entire Etheriean,Chanas, split, more or less randomly into different CHADS, which, as one ofthe big problems sory seeing that with like what loom network did with the DAPchain. That they'v recently tried to push out and I's, allowing adepth toscale on its own. You know quote Unquote Shard, while moving alongsidethe there of network. Is that HA correct way of looking at it? Yeahexactly, and I think, that's easier to and easier to do in some ways, becauseyou can sort of clearly define the boundary of your system, and you expectthat every interaction has cross chain. You kN W every interation in your atesinteerchain in every interactn between APS as cross chain, and that makesarchitectic of easier. But in the world, will you try and sort of just split themain atherium chain into different chunks? Then you end up with a sort ofweird you know: Combo World, where...

...sometimes you know other APS are on thesame chain, sometimes theyare in different chains, and you have tointeract them in different ways and so on. So I guess I think, that side, chains and so on are gonna. Youknow become practical sooner an, and I think that you know prooff stake isgoing to be a big component of making that possible, because it's easier intheory to write a m. You know something wet. It does cross chain proofs withpotent stake, because you know you've got finaloty, you know you're not goingo have to deal with Reyorks and so on than it is to make you know one chainalike kind of another when they're both Prut of work m. In terms of you knowsome of the other approaches I mean, I really think stake channels are goingto be the ones that GT. You know th. This are the first solution. We havesimply because m. You know for some well known, ewscases, they're, reallyreally supersimple and for other usecases, although they more complex,theyre, still TRACP, O Theroun releaset of unknown in there, and I think, we're Goinna see those youknow become a lot more successful and a lot more used when we start seeingin isdecays and- and you know, tall sets that are built for you know. Basically,here's your your Chaa, your state cennel construction case Um. So if Ihad to put me money on it, I'd say that I think we'll see wide deployment ostate channels before anything else, so you're. Looking at like the GeneralGeneral State Channes like I think what is Ame counterfactural group is puttingout and stuff like that. It seemed like an Thi Spang chain group, yeah stuff like that and mad Erians o. Whatsaying so so the counterfactual ones are really cool that they're a littlemore involved. I would be surprised AF before that we see, like somebody putout a really simple talkut for, like you know, the son has limosed Utilizy,you can N use. I, in certain field, you know certain applications, but if yourapplication fits thein, Haprey Stoe, you know pug this Libraryan and yourDan. You know, isn't that. I see I see first back upand kind of do even like a little more of a of a dive into different types ofstate channels. You have something like micro, radon, which has which was exxan.I gave you little explose at Devcon, three whereis a one way state channelin which one person iscending multiple small payments to a single receiver,and you have a full blown state channel which is two way and that two peopleput down a deposit. They then moved money back and forth ver very rapidlyoff chain, and then they settle that deposit. And then you have somethinglike Um Fun, fair spank chain, counterfactual Canterfacturalis,another ONEU. Or this moment you have the generalized State Channel and whichyou don't aren't just sending simple payments, but are updating generalizedstate of particular smart contracts that are offchain. And then you havesomething maybe like Counterfactura, which is Um. So all of the ones beforehand requiretwo on chain transactions, one to open up the channel Wone to close thechannel and something like counterfactural. If I understand itcorrectly only seems to require one, I think the in some cases even zero. Idon't know how that works but M and that you just open up the idea of achannel with somebody. You then PA pass signed messages back and forth as muchas you want off chain, and then, when you conclude your business, you basically post the settling transactiononchain, and these are the ideas of state channels which allows people todo. You know two people or potentially multiparty people to do as muchbusiness as they want, while minimizing the actual on chain transactions tht,all that all correct Yep and I think he pretty much described them in order ofascending complexity and usefulness. How about that NDSO? I and you se that, as as onescaling mechanism, how does that interplay into the otherones? ANDIT's a good question t for some? That will be all you need, asyou observe, with with counterfactual state channels. You can. You can havenot just one bit a series of interactions with only a single final transaction onchain m. So you could, you know, play, say a series of chess games or you know,a series of different sorts of games or business transactions or whatever andthensettle the aw with one final transaction and for that sortofApication, with at suitable R, the the single that might be all the scamability youneed that can support. You K an awful lot of users, but there areimplications where it doesn't fit. So well, for instance, where you have manyusers who, who interect more or less randomly with each other andso settingup individual channels between each pair of people, isn't really practicaland so then you're faced with either a sort of a hub, routed, nitwork similarto lightning and the eventual plans for...

Raden, or you have to go with anotherscaling solution like N, a side chain and the Mabie applications. Where youcan combine, say side chains and say chinnels say it say you have one reallyhigh volumeap, where even the the closing transactions a you, know morevolume than you really want to put on the main network an. But I think thatin the short term we likely to see people adopt one or the other simplybecause they don't need the additional complexity. Yet an you know, it's morelikely that we'll see those combinations in the long run, whereeven side channels start to get expensive. For instance, wone other aspect of side chanels thatyou kind of left out is it you lose data. You lose an audit trail and for alot of the people, I've talked to lot of the draw. That's coming with the UM.You know, Um things like a theory, Malanon waching in general s that youcould kind of keep an auto fiell, how you got from poinado point B and Wsigenas. You can take that completely off network and you lose that Deaa may lose that iyeah well yeah corus. Ifyou're not, then you may never be able to recover it wat if you're watching at the time,then you can record it. But if you miss e, then you may never recover it oncethe side channels closed or garbage cected or whatever, but you need to putthat history into the block chain and that's kind of the point we're tryingto. I think trying to avoid with a lot of these things that you don't want togo through, that that Rigomaro an also pay that storage fee, which is why Ikind of lean like okay yeah. These channels are great for now, but I'mstill kind of like thinking plazmais the next kind of way of going.Personally, I have I have kind of a Um J. I just think it's it's a good ideato experiment in plasma right now, Um. What are your thoughts on putting theens system into something more of a plasma like its own chain that connects to themainenace? Have you given thoughts to that or I I have a little um like I s,like I say nearly I kind of want to see how other applications evolved to usescaling first, so that we can serve the imbest Um from one point of view having your ownchain, thor enase would be pretty straightforward. A and- and you know, aplacema chain would be pretty much perfect and that's assuming you'redoing most of your resolution off chain, so you're doing it in Wallet, APS D andDepths with Weabon t faces and so forth, where it trips up abot t is, if youwant to do resolution on chain which is currently straightforward and andpretty affordable, gues wise or if you want to update things from inside. Youknow contracts that might live u outside the Enis World, then you haveto start doing. crosschain calls it gets acencronecent, expensive and so on. So I guess it comes down to in the end, the relative. You know the trade offsof how expensive it is it to keep it on on the main chain as opposed to a sidechain and how much you know how prevalent those usecases are ithink. Another way of saying that is.Who Cares? I mean Um wh, who needs to access thisinformation and what's the most efficient way of optimizing those people accessing it, I within speed and cost, and so,whenever you're designing whatever application you're doing. You need totake mostly those two things into account. Who am I interacting with, andwhat am I trusts associated with those people and the more you trust, thosepeople, the more you can move it to an offchained transaction. If the entirecommunity needs to access what you're doing, then it's going to be closer andcloser and closer to the main net? Yes, and in fact, M in terms of how complex this would bean the the route ENS contract. What we call the registry is very simple, so interms of practicalliy building this that wouldbe relatively straightforward to have a plasmachain that that chip poent to themain chain- and you wouldn't even you know- you'd- need the an the whole sortof appeall process, but you could execute any step of that Onchang inorder to verify it. So you've been come to an objective, ource of truth and soon, even though it may not be. You know, utx O, based, like the initial plesmaapplications very cool. So are you kind of a woneworld, black Cangu or an Internet of blockcangus? That's a good question. I I I thinkthat we're hitting in a direction where plasma, N and things like it are the the most obvious solutions. Sowe'll have sorry. Are you asking about scabailityfuture or e you to asking about maximalism maximolism s? I kindo wantto know like its like, like I brought it up really so thir's the OCAA. Youknow Pocadat and Cosmos, and all these things that are trying to you know swapbetween block chains are completely independent of each other. Logically-and you know, work have vthem all work...

...together in some way Um. I you knowyou're pretty focused on a theorium and M to get more philosophical here. Doyou believe that a single source of truth is kind of more important thanthe ability to have your own Um truth value and your own chain that kindointeracts with that yrouth value? Do you see the world adopting one backboneof truth and that being one world block chain, or do you see there being manyways of looking at value and connecting those ways will be the challenge? Iguess what you said about Internete at block chains, kind of Resen Anict, andthat I don't think we're going to have just one rock chain Um. I think we're going to have a lot forspecific appetations and they're going to be. You know all over the show, butjust like the Internet, I think theyill all be tied together in some fashion,or at least the bulk of them will. So we will probably have a chain of somesort whose main job is just to to. You know chip point state, Roote, Fhor, allthe other chains and allow them to interact with each other more easily,and you know less than a you ow, rather than a starthing. It may be that's the bactbone, but you know Youll, sere, sort ofpeering links between chains or it's relevant and so on, and that pretty much is the Internetactually because there's a whole series of smaller networks, t that are jointogether with a bacgbone, and you have these. You know offline networks thatjust stant intress and connecting to the Internet, and we look at it as onet against Ye. But really it's a pretty iherogeneous bunch of stuff, and on topof that, I think, what's important. To note is that the only reason that worksfor the Internet is because they all follow a single m kind of rough outlineprotocol and how they nd. Now they operate and move forward, and so what Ithink we'll see, at least in the being first implemented in terms ofInternet of block tains, will be all block tains that follow something alongthe lines of the e vm and, however, the VN moves forward, and so a lot ofthat's why you see a lot of the third generation or next quote unquote nextgeneration blockchains being ev incompatible because it becomes muchmore complicated, whin you try and cross talk between two very differentUm block chains in how they operate like thinking about Hatdo, the ethereumwould connect to Ayota or bit coin or something something. That's like a utxbase model is much more complicatev an issue than something like what definityis trying to create and that they're following the same basic protocol justdoing it on the back in differently yeah, just I actually. I did some didmy homework on your neck nd. So you had some a recent post on read. I, whereyou spoke a lot about the improvements you'd like to do on on eetherium Deing,protocoll itself and the EBM. Maybe you could go over some of those things. Youactually have a full list. Doyeah. I I guess from my point of view. Theriumsare pretty remarkable construction. It's got some really cool stub, youknow Checkin it, but it's also. It seems to me from looking at some of theespects of it that it was designed by snart people who knew their computerscience, but hace spent a lot of time doing practical engineering on on largescale systems. So there's things that make perfect sense from a seous pointof view, an an ECRIPTTO point of view, but you would have done differently ifyou had been familiar with the you know, the real world constraints and an thatsort of covers most of the things that I would you know in my imaginary,superitheorium change, so examples for that Ar, for instance, n the Midical Patricia Tree as I try andwhich is you know how everything stared on Disk Ou know, contract state andaccount balances and so on, and it is, I guess, you call it kind of spindlythat on Kno H, nod has relatively few subnodes and the result is that whenyou want to fetch a piece of data you have to go. You know many levels downthe try and you have to fetch. You know many into the GITANS from dusk. Each owish, which ae very small but real world discs don't operate like you know,imaginary, you know random exes memories. You Know Hard Dusks Have SikTimes, and you know ther bandith, this Hidh, that the Elagency is also veryhigh D and You have to seek from place to place, is, is Ds, don't have Sek,but they still have a trade off point between what the optiman size of lock tofeatures, and so an ideal dauter structure would take that intoma and-and it would have into it c ou much larger nodes. You know, or it wouldcombine multiple las nodes together into a single diskintity and that sortof thing could have led to a you know a significant performance improvement onreading and writing state updates it. It would be difficult, but notimpossible to sort of ritroat bat. Now you know, would be diffinently moresignificant change than any other ones who have had foutond so far, and there are other things like that,like choices around the structure of the evum and so on that again, you know,I see, is sound but didn't take into...

...account real hardware and most of thosespeak to peformance ipplications. Some of them speak to a sort of futureupratability and so forth. Yeah nd, one of the things you brought up that Ithought was something that's always kind of vex me is h when you're doingwhen I was I contract development, just the the the size of a single unit ofmemory is always two hundred an finty six pece and I was like yeah. That'sthat's. That's just not efficient storage at all, especially in storageheavy system like a block chain. It's like you're distributing your storageacross the network, and yet you choose the m. You know everything is twohundred and fifty six pecensose you can make it smaller, but it actually is nobenefit at all. To do that, it might even be a cost because you have to addsomething to check to, like rly, have opts to mniculate a smaller piece ofdata nd. So that's like to me. That seems the opposite of how you should do,but I'm sure there's choices, reasons they made those choices, but it's a good example because it's a veryattractive theoretically pure solution. You K W we're only going to have onebota type: It's going to be an ENTERJ. How big? Should it be well all of tour,addresses now keys, Asr, all Ofour, you know internal, addressing and so on andOu keys. Now, signatures are two fifty six bets. We ther need to be able todeal with two, fifty six but values. You know frfor accessing stuff. So whydon't we just use that everywhere in announced neat and tidy, but as you saythat comes Witha, you know real world practical trade offs in terms ofefficiency and not just storage pace but computing. You know if you want to addto you, know thirty two, but an just together y you're still going to haveto use big meth lie. You know o fide rooms because it doesn't know they're,not two sixpence right right, yeah and I feel, like, I think, a lot of thestuff that I've noticed in this early stage has been just lacking empathyfrom both an e enterprise standpoint, a user in herface standpoint, but thething that is really port. I W it gets the mess across and I feel like that'swhere theorum has been successful so far. Is it it? It actually makesprogress m. They just did it, they might have not done everything optimal,but it got the job done and now people are talkn. Now people are exploringthese technologies and I can actually deploy smart contracts and you K ow.Maybe it's maybe the rushed nature of the I call it rushed. Maybe they call Ithoughtful, but the nature of it Um seems seems to work out in the long runanyway, because Um, you know, there's success. There's definitely marketsuccess in what a theory is doing speaking of Success M. Do you think theNS will betogonized at any point? I don't think so. Um I mean not nft. I mean like Liear S. twenty,like is aglacy yeah, I mean disclaimen here you know. I said thatwere we're really looking for people who have some system design ideas. Ifsomebody came along and said, look with a FOUNDOF o an o solve all of yourproblems. You know to to make rent, fair and- and you know, have Hav asystem that you know allocates names farly and and makes sure that theprices are riht and so on, but to do that, if to she tokens I'd, be likefine Welisho takens, and but we ere only going to do it if it's like anaircet to the system, if it's something that fundamentally makes it operatebetter an if it's just you know, get that tthere's, no real reason to IssouTokens Gotya at O, so nl answer go ahead. Cort such that was such anengineering answer. I don't if the technology fits then sure if it doesn't,and why would I use it exactly tats a good business answer too. I meanyou, don't want to token o seem to be adding Ale bit of Risk Gellement, sowhatever's going on right now we don't know where they're going to go. Wedon't know how they're going to be represented. We don't know how theycound be treated, whether or no it's the utility token or or not it's alittle difficult to ascertain. What you know are. Are you printing money bydoing that? Is that something that certain governments would prout upon,which ones there's a lot of the questions in gray areas surroundingthat? So I mean, I think, it's wise from all fronce, but yeah. It's prettyengineering answer, so speaking of which, like these areall UM. All of these H, these these names thatyou're stick up ere these domains, they're going to be represented by theseven twenty one or the Um. A E thin nft is e. seen tw on or seven twentyfive Abrain doesn't want to recall that I thing yeah sevenieat or the other. Ite SA Yeah I've, given it a little thought. It probably makes sense to make the newregister ou now es seven twenty one compatible just because people are developing likeGenerio size. You know auction contracts and you know other ways tointeract with and manage those resources N. bearing in mind, of course,that's only going to be your. Your second live all Doddeth names, so thatdoesn't mean the whole system. You know every name would become a sin tweny onetoken, because the base system is as a much simpler API than that and, ofcourse, was written before seven to...

...anyone existed. So you know definitely something Polcould empomanting, but I believe for the DOTIT registraar. Do you think you'migrate to a seven twenty one system? It seems like it's being adoptedquickly as a standard. Thank you, hello, kity's, by the way, CREPTANWHOO's ISO. So there's one componiont to theUNISE, the registry, which would be very difficult to upgrade um in that we would have to copy overevery record and t en tell people to use the new one. You know be more orless manual, and that was your know. Intentional design we'll have this onecomponent, which is your like sort of indirection layer, and we make it assimple as possible, and then you know that points to all the upradable birdsm. So if we really wanted like everything to be a seven twenty onetoken, we would have to replace that and I don't think it's likely simplybecause I don't think the value arenants it. The current contract isabout a hundred lines long and if you delete the comments it about fiftylines, long, there's no loops, there's, basically, no, if statements it's aboutas simple and as as trivially ardered bousigable as you could possibly hope-and I be very reluctant to compromise that by by in complexity, unless wecould show like a really compelling usecase for it Um. So, okay, that makes sense yeah, Ikind of feel like you would treat it Domaina same way. You'd kind o treat atoken n that you'd want to transfer it easily between one party and another,without a lot of barriers and since there's already standard interface forthat with the nft standard, it seemed to me that, logically, from a userinterface s tanppoint, it would actually make sense that you couldintegrate tool sets which also adopts nft. So the good thing is contracts can ownnames, and that means that you could write. You know seven, twenty one, youknow Rappep, for instance, and that owns the name and then can deal with.You know, tokenizing it that makes perfect sense yeah. So whatabout identity? Do you feel like the identity standard is gettinganattraction? Maybe some ky C elements of what you're doing with the e Ns some yeah. I I haven't heard a lot ofpeople. You know working on. I did see in talking bellionus, but then mostlyifforts around dying to see that I've seen have been keeping feerly quiet angeneralso. So I don't know what they're using I diff. It's definitely somethingI'd like to see Morof, because I think that Tian idense to are name asfonomeal component. The whole thing- and I think ineuss is well placed to doit because, certainly with the Doteth toplivl domain at least you can havesonstead of guarantees. You need about control over that name. So you lookedinto like H, Yar c eight, twenty five e one of the Fabian Fobal cellersproposed for for identity at all, or only at a very high levil. I I I'm aneat editor and I try and keep up on stuff. But often I like skin the original proposal and read thecomments and tean Ti Ol thetime, because I just don't have time to readthem all. Yeah Yeah. I'm the same way: Yeah Sonly, especially when to getthree white papers every day, Yehm like one of the pushes to abstract, addresses on isoom network.How does that Planto, the NS system and and first OFFCAN, you give an overview oflike the what it means to abstract addresses away and then how itwould plantnus m. So I guess en eastern General is anattempt to abstract addresses away, but I I think maybe you're talking aboutthe Metrobalis hardfork sugguct jom account cobstraction, so in accountofstraction you still have addresses, but you no longer have the samedistinction between external accounts and contract accounts. You would beable to provide source code for just abotany any account and that sourcecode, wood or Bake hode would determine how signatures were valigated and howfees were paved and so forth. And so the idea is, you take a bunch of stuff.That's currently hard coded into the base of the ETHERIAN system likechicken noncers and paying minning fees, and you would turn that into you knowflexible stuff that can be done through contracts and it was originally plannedfor what became Bozentium, but it became apparent that thee'. It changesa few invariants and system that that people have been unconiously relying oncouldn't need to be a pretty big warning period and plenty of time forpeople to update their expectations. So one example of that is worth account. obstraction. You couldwrite a an account and have a transaction that gets mind multipletimes, so you could heave, for instance, in a lanb clock account that anyone canstmerchants. You know and run the alarm clock and it gets paid for by thepeople that you know wanted their contracts triggered and that sametransaction. It wouldn't have anos and the same way as current one does couldbe submitted every block or every tenth...

...o'clock or whatever, and it would havethe same transaction Hash, but it would v executed a hundred or a thousandtimes and lots and lots of Cote. In fact, pretty much all the code at atthe moment treats a transaction hash as a unique identifier for the ectualexecution, not just the contents of that transaction. So it would break alot of stuff if suddenly it could be executed, five or teen or a hundredtimes. Isn't that doesn't that also haveimplications on who pays for what insted like right? Now, it's it's.Whoever sends a transaction pays for it. Doesn't this change also make it sothat the receiver pays the gas fees associated with transaction? Yes, sothat's the that's one of the big usecases, for it actually is being ableto build systems where somebody else pays the guest, so you know make iteasier for users to interact with APS and soon n in controlled, fashions Um.The question, of course, is how a minand knows whether it's going to getpaid at all when it excutes transaction M and there's basically a couple ofsuggestions around that the simpluce Todos that each miner would have awhite list of Bikego of contract bike codes. That knows a trustworthy,basically, that no nose w will pay its for a share and then, if you wa to launge a newtype of Accountw, you would have to you know, standardize that code and thennconlutes the miners. You know that it's safe and they can turn it on.Basically- and there are you know, more complex suggestions, but they theyultimately an you know. The miner needs some way to cheaply determine whetherwhether it's getting et paid, it sounds like a credit, score sort of but kindof yeah little yeah. The issue, of course, is that since the contractsthat you're in conplate, you can behave abateurily sneakirly and you culdn, youknow, write something that behaves well for a while and then stops paying theminons. You know. So it seems that the only opctions really are either to headsome way for the contract to pay the minor at the beginning in a way thatcan't be n undone with the reverse or whatever or some way for the minors tolike o know for humans to go in and assist, thon and and choose which onear safe, O. Well. Couldn't couldn't be part of th, basically pre prepay, theProesso, basically you prepay and you can actually auto funds. Whatever is inyour bank account for the payment kind of, like you know, prepaid cellphones.You would basically contreat it the same way. You have to pay this much meminutes are up. No, we stoppd doing work for you lately, you could. Youknow the minor could run say the first thousand gas with the transactions andsee whether they prepay o wito instructions and pr see if they'eprepaid by the time they get there and if they hear if they keep on executing.But of course you need some way to make sure that prepay can never be. You knowreverted, so it might need some changes, O he bm to support Er. I think it's very cool that it there'sto say it's going to a change, ane assumption so to actually get it rolledout. We're going to have to you know, soledify how it's going towork and t give the community lot of time to update I'm pretty sure mannetyif the skin's going to be pretty unhappy. So what is your? What is yourthoughts on upgrade pats for for these protocols? IMEAN? Obviously, a lot ofwhat you said Um throughout this this you know, interview has has evolvedsome some lacking things in both the evm and just a protoco itself. What what? Whatdo you see is being so? How are we going to upgrade? Ishardfork sufficient? Is there some some other mechanism like it was tezos thatthat they had some sort of way of upgrading, but it was pretty muchcentralized Um, I I is there. Was it te o? Is there some sort of like what doyou see being the PAS Opgrad ability on the network, I'm I'm kidef withfatalicanplayed on this. I think that fourking is actually an ASEASIN USregard and and having governence off chain as an Airsetionis regard, because its Wa gives Blok change their power.EFVICTIVELY is, is that their consents consents US and consent BHASE systems n.You can't have a majority Bullya minority into excepting a chain changeto the chain, because the minority can fork off and create their own chainwith blackchack and hookers B. Basically, that you know stacis isthe default and that you can build your own. You know your own variant, but youhave to convince people to come along effectedly and the problem withonchained governor systems that assumed that the majority should automaticallybe able to do. The thing they want is that they they take away thatopportunity, or at least they make it a lot more difficult. Do you think ie ageis sufficient, forcing upgrades, I think it is yeah. It's I mean one wayor another. Anyone who wants O to keep going withwithout Y Kno, a givenupgrade, will at least have to Falk to change the Ice Age. So I think it doesa good job of of ensuring that Staus a arn option, but notindefinitely so...

...you're goingto have to make a choice dto to make a change. Even though you'rethe one you know you get to choose what that changes cool. Well, I I just have one morequestion before I handed off to th cort o wrapping up what project are you kindo most excited about n at Eram Space Right now? That's a really good question: Yeatreally whyt, like I really want to know. L K what's out there that I haven'tseen, maybe something that is kind of on the fringe that most you've heardabout, but it's not really gotten a lot ofattraction. Yet mmputting me on thespot here, then I ou d say that Um th, most of the ones that come to mind,are ones t at that people are pretty aware of. Like I, I really like whatAragonare doing I like their whole renical transparency, tutransparencystick and I like t they're building likepractical tools for for real world companies to to use to, you know, bemore transparred about thei governance and so on. Um I I know you know Bankwarr a bit sortof o kN ragged on and an fill disclosure. I I ordered their contractand I've you knowtethey've consulted with me occasionally, but I think thatthe attakers is pretty neat. You know in that the idea of being ableto do like an instant on chain, synchros transaction to get. You knowchange between any two coins. As Pretty Neat Um. I guess the stuff that excites me mostat the moment is actually some of the scaling efforts which aren'tnecessarily you know, big ICOS or whatever they're, just research efforts,Um and some of the other stuff that thatexcites me and sort of fhither off or more hypothetical, like I would love tosee. somebly build a random number DAO based around bielis random beacons. I I you K, ow love to see some of Theecasnak stuff that is now possible, a giving better language, supportn andbeing more widely deployed M. it's yeah, I guess one of the other things I Iwanted an excuse to talk. Wathout in fact was Um stateless clients. Vlades is really not a fan, but I thinkit as Al Os potentials there to build a assystem, ond scales, better, simply'cause. It doesn't require over and to stale the stuff you know. Can you canyou explain that a little more? So the really short version is that M.instead of actually recording the transactions on the chain, you M, yousubmit the transaction, but you only keep track of the final state and you store all the individual datafor your air locally. And so basically, everyone who wants to use a particularappes to you know fetch that data from you or they have to watch the chain toget that data. But individual nodes. Don't need that Dotar in order to beable to verify that a transaction as valid, because each transactionactually includes that it's IAONIN put data and proof that that is part of thedata see it. So if you take, for instance, a token contract as a sit ofbalances and so on, the users of their token contract to the only ones thathave to store a complete copy. And if I want to transfer team tokens from me toyou, all I have to do is seen t e transaction saying transsfit in Tokens,and here is the proof of my current account. Balance using like a mirkleproof, was Tani an nods ware, able to verify that there was the differencebetween that and a you know. Complex event system that we already have onatherium and fo the events, let you prove a to alight client or a fill client that something happened and in what thedetails were, but this let you actually enade all effectively for clients thatdon't have to stall the DAUTA in the first place so woul this. This sounds to me likeyou're, also you're, sending additional data that'swell, I guess technically,would always be sent Ke. Does this increase network useutility of of the protocol at all or is h, haws usage of the protocol alr likeslike? How would that impact things gloi it would. It would increase the networkutilization because, as you say, you'd have to carry along the proof. Darte the transactions get bigger, but thechain gets maller. I could be less en issue going for itin the future. Yeah all right. I think that's a that's agreat way to wrap it up and I always kind of enjoy hearing what people umare excited about. It gives you a really good overview of where theirfocus is based on what they're looking forward to and all of your answers werebasically all infrastructure level, which tells you kind of where you thinkwe are as a whole amongst the entire technology and what we need to moveforward and from what I clean from what you said. We need more infrastructureand Um. I want to say nomost platform typethings that enable people to build things. We're not too. We don't caretoo much about what the in user is seeing in thosesites of applications relative to the things that are enabling people tobuild those things in the first place.

I I definitely agree we need moreinfrastructure. I wouldn't read too much into my focus on infrastructure,because what I ass about, regardless of Amatur of the pletformance, that's whatI love working on yeah, I I I mean I guess I am aligned with that with thatviewpoint. So that's what I saw from it, but h things for coming on the showneck. I I appreciate all the work that you've done: Um for the entire spaceand you keep doing it and anythinks appreciatit can.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (108)