Today Verkle + Tomorrow ZK = Everything Stateless, Everything Lightclient | Devcon SEA

[Music] [Music] hey everyone our talk is called today veral plus tomorrow ZK equals everything stateless everything like client so my name is Jason Chasin I do various different research projects at the ethereum foundation and in the first half of the talk I'm going to give a highlevel kind of beginner to intermediate overview of the benefit of stateless clients and what they are and I'm gajender and I'm a code developer and I do a little bit of research as well so we'll I'll talk about the endgame of ZK ethereum yeah so before I go into explaining kind of what stateless clients are and how they work I think it's worth it to give a quick shout out to the stateless clients teams and everyone else that's researching stateless clients because I'm just somebody that tries to explain them simply but without the hard work of people like Yom who just spoke ignasio Josh and so many other people like none of this would be possible so maybe give a quick Round of Applause for [Applause] them okay so there's some common misconceptions that um if you want to run a node you're just doing it for alteristic reasons or that you need 32 eth and none of those two thing neither of those two things are true there's other reasons including getting trustless information about the blockchain maybe you want to know uh your balance of eth and you don't want to trust in fur or centralized RPC you can run no yourself and um on top of instead of asking them um since they they don't have a record of all your asks now you have that private lookup access and uh we know what web 2 companies do with um your data so we prefer to keep that private and uh you also have trustless broad um broadcasting of your transactions you have direct access to the P2P network if you need to broadcast your transaction and of course it strengthens the network decentralization so for all these reasons I wanted to run a node about 8 months ago and also um I felt at at well before we go there who actually does run a node here please raise your hand okay about a third of the room and um I went to a couple italic talks and he loves to ask this question and I never had my hand raised and I felt left out and for that reason and the ones I just mentioned I really wanted to run a node so about eight months ago I posted on farcaster and I tagged Bounty Caster because I'm a hardware Noob and I needed help setting it up so I asked for someone's help to help me run a third um to walk me through running a full node and uh thank you to limes who's actually in the crowd here for telling me what Hardware to buy and how to set it up I I hope you didn't spend that eth because it's worth more now and so the first step I did is I went and I bought the hardware I spent 600 bucks to get a new computer and around 150 a little less to get more storage because you need some more storage to run an ethereum node total cost about $740 and then I actually set up the node like I said I'm a hard boarder Noob so I kept copy and pasting things into chat gbt this process was a lot more confusing than I thought it would be it took a couple hours and I was pretty pumped once I finally got my node syncing so I took that screenshot there and then I left my computer and went back a day later and my node was finally live processing blocks um totally synced with the network and this whole process was a pain I definitely understand why not all the hands were raised it's um you do get benefits but it's it's a really pain to set up the node now I'm going to walk you quickly through why this process took so long it took over 24 hours to get live and first it's because you actually need to get all of the previous blocks and this process takes time if you're in blockchains big um you you not don't necessarily need them all but you need at least some of the blocks and um you also have to load up the current state the current state is the current um amount of eth that all accounts on the network has as well as the data that's stored in smart contracts so you need to know that I have two e g has 5 eth you over there in the crowd have six eth and then you also need to know the current state of Unis swap a and all of our favorite smart contracts and um this process takes a long time you have to go into that storage that I told you that I had to buy later and update everything to make sure it's live and that's why it really takes over 24 hours so now once the blockchain is live and you're sync to the network what happens when you're verifying blocks well in the first couple seconds you receive the block from whoever is the random chose randomly chosen block proposer and then you do the very simple checks you check to make sure all the signatures are valid you make sure it's the correct po proposer all the header stuff is correct and then you actually have to go through transaction by transaction and each one access your node um State and see okay if Alice sends Bob one eth um does Alice in fact have enough eth to send over and then you have to update the amount so subtract one eth from Alice add one to Bob and you have to do that for every single transaction that is in the block and there's something called the state growth bottleneck of all blockchains have to suffer through this and as the blockchain gets bigger and bigger and bigger it takes longer to actually look things up in the state so in ethereum in day one it's easy just add on one account and update it one e but once the note is really and once the state is really big you have to go find that one account you need to update and then um that process keeps getting bigger as the state gets bigger and I'm going to hammer that home with an analogy in the next slide and the processing time takes around 300 to 600 milliseconds per block in the last 8 seconds there's some consensus voting stuff that we don't need to worry about all right now let's Hammer home the state growth bottleneck soag um first let's take a look at that graph by Paradigm it's it's really awesome they're producing really great blog posts going over how we can raise the gas limit and as you can see it's up and to the right for both history and State uh We've invented up only technology just uh not in a great way because you have to keep updating your storage as this gets bigger and bigger and um now let's go through the state growth bottleneck problem so imagine that I'm a mailman in New York in 1800 and um I have to deliver three letters in 1800 New York's not that big maybe it takes I don't know who knows how long it's not it's a small City you deliver all the mail now you have that same three letters but imagine you're delivering them in New York today the city is massive who knows how long it's going to take you to deliver that so that kind of shows even though it's the same three letters it just takes longer as the city gets bigger ethereum blocks are the same way as a state gets bigger it takes longer to actually go through the transaction list and update um and update the amounts now let's talk about running a node in a stateless world you don't need to buy any new hardware just use your computer that you already have and uh you don't need to buy new storage your your storage is good you actually barely need any storage you install choose and install your clients the same way and boom you're near instantly live you don't have to wait 24 hours click your That Was Easy Button if they still make those I'm going to quickly walk you through technically how it works um somebody who builds the block does still need to hold the state so either that's a builder maybe it's a soloer that chooses to keep um the state because they want to build the blocks themselves and when they create this block they also attach what's called a witness it's called an execution witness now this witness they actually create a proof of all the state that's touched and and in this proof it cryptographically 100% proves that all the state that's touched is in fact part of ethereum's current state so in my example I keep using of Alice sending Bob one e it might say Alice 100% does have um let's say their Sergeant balance is two eth and it's sending to Bob who has five eth the proof will prove both Alice's balance of two eth and Bob's balance of five e and then your stateless node you receive this block and the witness and you don't need to look up anything to confirm it anymore and you could actually do the entire block um processing in Ram which is an extremely fast think of it like a trackpad um where you can just cookly jot things down and so this this um process is a lot more efficient because you don't have to go into your state and look things up anymore more one thing I do want to mention is these proofs need to be compact um this wouldn't be possible with Merkel proofs just because they're too big so we've slayed the sa growth bottleneck the the L1 will actually be able to process blocks 3 to 10x more efficient and yes this means we're scaling L1 gender is going to talk about scaling L1 even more and uh if anybody tries to FUD and say ethereum's not scaling L1 just send them a link to this presentation because that's not true but I'm just getting started one thing that's super exciting is we're going to actually be able to build these stateless nodes into your wallet so because they sync instantly and because you don't need any storage you're going to pull open your metamask and in the background instead of connecting to infura you actually connect to a stateless node and you get all the almost all the benefits that I mentioned earlier and you don't even realize it so this is definitely going to be possible in your browser and like possibly your mobile phone so people disagree on whether it's going to be possible to run in your mobile phone marus uh from G he thinks it might be possible and he says ethereum can grow from 10,000 nodes to millions of nodes danrad also thinks it could be possible in your mobile phone however ignasio told me that phones and SmartWatches may be a stretch however in the endgame ZK we'll definitely be able to run them on your phone and your SmartWatches and I want to talk about how important it is that we're going to have like maybe millions of ethereum nodes this quote by ballic really stands out to me it's pretty much saying that like if doesn't matter how many nodes you have if they're all in the same group chat maybe the foundation pays for people to run nodes and you're not verifying it yourself you could wake up tomorrow and see that the foundation now has or you won't actually see it because you're not running a node but they just print more tokens for the foundation and people don't catch it until a couple days later at that point it's too late they come up with some BS excuse that they need it for research or whatever but that's just not ethereum way and it's not it's not good enough and I think ethereum is like real inspiring in this mission to make sure every every body runs a node whether they know it or not ethereum isn't just a tech startup it's like a global public good for all of humanity and I just find that so inspiring and motivating and it really fires me up so maybe let's let's cheer for [Applause] that so quick recap with stateless nodes you're barely going to need any more storage the verification time is at least 3x faster the setup is near instant definitely will work with your browser mobile device eh uh number of verifying nodes potentially Millions but wait there's more advantages we're actually going to be scaling L1 L2 because of this the data layer scales um it's going to scale because of this thing called Data availability sampling and the more nodes that there are in the network the more we can scale the data availability layer and remember how I said there could be millions of nodes yeah you do the math there um also they'll be potentially able to be better cross L2 communication because these proofs are going to be very small we'll be able to easily um send them to the bridge and verify that certain things are part of the L2 State and it'll be easier to do um new constructions where the L2 Bridges communicate with each other and the last thing is in the witness you know all the state that's being accessed so maybe you decide you only want to process blocks if they touch State XYZ or whatever then you can do that because you have that witness and it shows you what part of the state is being access and now I'm going to kick it to gajinder to talk about the future all right [Applause] so seems like statelessness and work I mean this this sound like magic we should really do it and uh maybe with ZK enabled ethereum we'll be able to push the boundary even further so the end game of ZK ethereum can be seen through lens of State compute history availability finality but uh I'm sure that everyone remembers scalability trial so this trial is like uh you know classical physics you know it so it most of the work most of the time it holds but uh then like uh quantum mechanics you can push the boundaries with some magic math and as Jason has told you shown you that while when you run a road node you need of space and although most of the space is not used by state only 15 to 20% is used by State uh but even if we get rid of that and uh but we would really want Giga gas and that would basically explode the straight so L2 sort of are a shed straight and ZK rollups basically approximate that Shing a little bit well uh but we also need engineering solutions to Shard slice and prove the straight and for that we need statelessness and one can think of it like that almost complete removal of State without compromising security so the the other big thing that is on the El clients is the history and uh even though that history is not really complete we need eths logs and uh but there are ways we can do less and shorted history EIP for 4444 solves it a little bit but what we need is ZK friendly accumulators and filters for Rapid Global indexing and lookup and obviously engineering Solutions and products to short slice and proof the history for computation we can do distributed block building with ZK EVMS with verifiable straight transition uh maybe we should also enable huge smart contracts like the way wal does with a unified tree provable Cod chunk on demand loading for availability we have blobs we are pushing shed blobs basically jointly verified dat availability and maybe we should also do shardy transaction meol finality uh we have AIT but we can push it even further with having multiple committees with ZK verifiability easy ZK aggregation which gives us fast ZK verifiability and could lead us to have faster slope times so we would want to see the protocol as some of it slight client parts and all these are composable but provable islands of State compute history availability finality and the magic Source behind this could be stat lessness and ZK and the Common Thread behind that is polinomial because polinomial are ZK friendly constructs that can encode high amount of data and commit them into small size valuations so State history computation trace it can be mapped onto polinomial and generate one can generate suent proofs in size and verifiability but there is a downside that it's a little bit resource in intensive uh but there are some strides that have been have been made recently Circle stocks miners but we believe that uh the further research into it will unlock more performance so with statelessness so we can conclude that you know not all participants are going to be same in the network not all participants will have to do all or same work work done by one can be used by many there is some centralization risk but it is mitigated by easy verifiability and highly decentralized attestation Network that can be built on top of it and we can have shed engineering solutions to to basically mitigate some of the centralization L where I think portal Network could play a big game over here so we might go into an era where no one is a full node or a participant but all this composability needs to be protocoled in but I mean we there is a lot of hard work hard work that is still ahead and a lot of research that is still ahead so we should be a bit cautious in our excitement but anyway I mean I think ethereum can be really can provide this really exciting ZK future to its users where you can have fast transactions you can have smaller clients nodes and basically you can have bridges that are light clients and they can verify and you don't really have to have full node running so I guess you know there is exciting times ahead for sure [Applause] thank you very much for the wonderful talk G jender and Jason so turning our attention to questions in a rollup Centric World a lot of the status and blobs and users are interacting with rollups rather than the L1 can you talk about statelessness in the context of L2 sequencing so uh as I said you know uh ZK based rollups are essentially a very good approximation to a Char state that you know they they have the data that they post to the L1 smart contract they post verify they post proof of that they post uh proof of State transition and that gets accepted so in in a manner they they are like a shed extension of L1 although with ZK VMS you can even you don't even need a blockchain sort of execution environment you can basically have uh any kind of data structure that you want you can commit to it you can have your whatever is your logic that you update that data structure with you can do that and I think you don't need even an L2 for it how can stateless clients be used to increase da if running on phones or laptops with limited power that are not always online and they have no stake to slash if they withhold data well this is just this is just sampling um so it's not like they're proposing blocks in this world like I mentioned if you want to propose a block you still are going to need the state these are just additional additional nodes that are sampling the data so basically what essentially uh one can do is that you know you have these small nodes which can easily come up using light client protocol and then you can uh get that data and build for example fraud proofs if they are there or you can actually verify whe if for example there is a ZK proof of State transition you can verify that uh in a ZK uh State transition based world this would not be an issue because smart contract will not accept it uh but if it does I mean that is sort of a bug and I think the social L will take care of it uh what trade-offs do we have with statelessness so the trade-off with statelessness is against the resource that is out there for example if you have seen in one of the memes uh made by GM actually that uh if everyone is statelessness how will you build the blocks so there is particularly you know you have to rely on Builders out there we who have full state or you have to rely on uh decentralized infrastructure provider like portal to get the state and uh then you will have to build it so there is some centralization risk but we believe that uh highly decentralized attestation network is actually more important and we can make sure that if the cost to block building isn't too high then at any point if there is a risk any kind any number of Builders can enter the ecosystem and start building and we're already actively working on like distributed block building methods so if you want to follow along on ethereum's um YouTube page there's already been a breakout call for distributed block building uh stateless clients exist uh what's the incentive for normies to run full clients uh so in my vision or me wouldn't run full client they would for example interact with something like portal uh they might still process transactions get it get the state from over there uh give it to uh submit it to the transaction of the nodes and they will uh do their consensus rules and make it part of the blockchain uh but but again you know the incentive is block building for sure if they want to run a full note yeah and that's also kind of part of the point that I was really trying to hammer home um they're going to be running these nodes kind of whether they even know it or like it or not so the incentive doesn't matter as long as they're built into the wallet they're going to be running it uh how are conflicting concurrent transactions going to be resolved so I mean this is more a research oriented problem and in my believe Yeah so basically when you apply a transaction when when there is a stateless transaction that you will apply either if it accesses a state that has been modified or if it's accessing a new state because uh smart contract execution is touring complete and if for example dynamically if there is a change in value it could uh totally access another state uh if that happens you will not include it as a part of your block and that transaction will be resubmitted again unless you are a full node and basically can access that state or get it from the portal CL uh from the portal Network and build the block when gam all right great uh what's your opinion about people claiming that having 32 e means validator validators have money to buy more Hardware you don't need 32 e right now we even with 8 or even with 1.5 E I don't want to Shell the protocols out there but you can run out validator and your uh reward will not match the cost of running the hardware so what's the role of l2s in a veral plus blob or ZK future L1 the same thing they are today um we want more space more um transactions where you're inheriting ethereum security okay and does this work only with Das what if committee doesn't distribute the block over the network what consensus modifications does it require uh if uh committee doesn't distribute the blog then the blog never gets imported because uh the the consensus clients don't import it it never gets part of the chain there is data availability checks and this is further even enforced by data sampling that even if you have shed data that you have uh sort of you have imported and because you want to F date and then there is whole bunch of other data that is out there that you are not sure whether it's uh available or not but then uh for example P runs a sampling protocol and full Das will also run a sampling protocol on top of it to make sure that the peer that you have they have uh the data that they claim they have and uh on the whole uh if you look at the network there's some probability that they they might not have other data that uh they say I mean they have imported the block but they might not have other data so there is some probability but if but if there is a big Network and if you have a good pairing And if every honest validator or beon client enforces uh this sampling Network as a whole will have the data all right well that's it for questions thank you very much again uh