Next Generation Based Rollups: A Practical Approach to Unifying Ethereum by mteam | Devcon SEA

hello Dev G all right so we're going to talk about next Generation Base rups today there are some base rups that are Next Generation there are some base rups that are last generation you guys just learned about some um I'm going to tell you more about how you can become a Next Generation based rollup kind of what the definition is and then a little bit more about how I see the future of Bas rollups evolving cool so this first section we're just going to go through some of the traditional base rollup research talk about the the Bas rollups that have been you know explored in the past some of the designs that have been proposed um and then we're going to understand some of their flaws some of their weaknesses so one of the very first designs that is used by Tao today on Main Nets is total Anarchy base sequencing this is like kind of letter of the law exactly what Justin Drake proposed in last year right and it's it's very simple very basic right it's called total Anarchy because it uses a uh election system that is permissionless um and allows anyone that's the Anarchy bet to propose block so one key Point here is that the sequencer Is Not Elected beforehand you don't know who the sequencer will be beforehand you might know the ethereum proposer because you uh can look at the ethereum look ahead we know 32 in advance uh but the sequencer of the base rollup might not be the proposer teo's case it is Techo labs for the majority of blocks um one other thing to note is that layer two block proposing is completely permissionless right there's no permissions involved uh the way that this ends up playing out in practice is that we have this diagram up here uh we have a bunch of sequencers proposers in Tao's system that give their blocks to the Builder to the me boost Builder that would be Titan winner mute Etc um and they would take these blocks order them in the layer one block so you know each block proposal is a layer one transaction and then send this completed block to the proposer through a relay and through the me boost auction right and then eventually the proposer would propose this to ethereum once they've selected the block with the highest bid so it's up to the proposer and and the Builder to choose which block created by these uh layer 2 sequencers or proposers actually gets onto the chain now one design that aims to solve a few of the problems of the previous design is vanilla based sequencing so lime chain um George put out this excellent bit of research that explained how we can actually solve a few of the foundational problems with total Anarchy baseding right the Teo model so the the first kind of thing that we we try with vanilla Bas sequencing is a simple sequencer election right that gives us the ability to choose a sequencer based on some rule we have a primary Rule and then we have a fallback as well so the fallback in vanilla based sequencing case is just total Anarchy so there's still some total Anarchy here uh and the primary sequencing is based on some election mechanism that is not clearly defined and kind of left implementation in the initial research so the other thing that vanilla Bas sequencing includes is kind of very basic primitive support for delegation which is allowing the ethereum proposer to give the rights to their base troll out block to an external Builder without directly going through Mev boost is like external to the Mev boost auction vanilla Bas sequencing also included basic support for pre-confirmation which gives the you know we heard about it last talk but gives the uh proposer the ability to promise things about the block that they will eventually propose um it does require some sort of Delegation or some um increased Hardware constraints on the proposer and so the design for pre confirmations has kind of wide reaching implications for centralization vectors for for stake and also for uh pre-confirmation in real time now the other thing that vanilla Bas sequencing introduced that really caught my eye was Revenue generation from a percent of congestion fees and you you'll hear Justin Dr talk about this all the time you'll hear me talk about this base rups can capture revenue from congestion trivially right these base fees if you're doing EIP 1559 evm right they are represented on chain and are kind of understandable from within the the evm within the state transition function so you can view things about the the fees you can take these and um you can do things with them right you can direct them to a treasury you give some some of them to the proposer or you could return them back to users so kind of the the problems with these that are actually really big problems like Tao on mayet live is extremely inefficient um all Forks of Tao without major changes will continue to be extremely inefficient um one other problem with kind of total Anarchy and vanilla base sequencing is that there's kind of no built-in composability any composability is on a layer above the the sequencing and that's really unfortunate because whenever you have to build kind of another layer you introduce complexity and centralization vectors of some sort in the case of Base rollups that's quite possibly around pre-confirmation which means we could have centralized ethereum stake which we all want to avoid other thing is M Revenue so lots of people still have this misconception that based rollups leak meev revenue and that's true for traditional based Roll-Ups because they do not use um an auction of any sort to kind of get an oracle into me so if we look into layer one me research we'll see things like execution tickets and execution auctions um that are all designed to get some Oracle price essentially of the me in a block now the other thing to note about Bops like I mentioned before is inefficiency so Tao has a about a $10,000 a day cost to to do their sequencing um before blobs they had about $100,000 a day this is a lot of money clearly and the kind of thing that you need a massive amount of users and network effects to actually be sustain aable right and that's something we we obviously want to avoid but there is a solution for some of these uh these Tao Forks we've been looking at doing slower block time so Tao now has a block time of 33 seconds instead of 12 seconds um which is of course ethereum block time that is you know that has allowed their cost to Cal down by about a third which is good for them um but it's also really bad for users you get a worse user experience so the way we can patch over this is with pre confirmations just throw some pre-con on top uh the downside of that is that without doing your pre-confirmation designs in a clever way you're going to end up with limited benefits so your users are going to get very limited availability of pre- confirmations um and of course a limited composability so unless you're using a a shared pre-confirmation protocol you're not going to have composability so we start breaking a lot of things when we add pre- confirmations uh the other thing that that you know specifically Taco's case you lose composability with layer one because pre-confirmation happen you know over a course of an entire Epoch perhaps and during that time layer one state could change a ton and so the kind of current layer one proposer is very different than who's actually sequencing the rollup that's not sh sequencing that's not based but we can do better traditional designs are not good enough we have over the past two years we have new research we have new teams we have a lot more data about Bas rollups because we have one live and we have a lot better technology so one of the big things that will be talking about in their next section is ZK um using snarks to save costs and improve efficiency so let's kind of Step take a step back let's think from first principles with all the data with all the research um and with all the the the teams that we have now kind of what are the things that we really want to get out of Base sequencing the first thing is in in my mind the most important which is synchronous Atomic kubos ability and you know this is a a buzz word for sure what I think that means is atomic cross chain contract calls that are sufficient enough for Flash loans and the reason for that is because Flash loans are a foundational tool for doing Atomic Arbitrage between domains so once you have cross train contract calls you can do Arbitrage um you can do flashlands for Arbitrage and this gives you the ability to essentially Arbitrage prices on different domains giving users kind of the best prices uh across chains you can also use that for doing like a swap that uses multiple chains liquidity if I'm on arbitrum I could directly use ethereum layer one liquidity without waiting for like a a service provider um or you know some collateral now that's pretty straightforward the other thing to to look at is a a seamless user experience and developer experience so anyone who's ever tried Mak an asynchronous application will know that it's 10 times easier I'm not joking to build a synchronous application if it's asynchronous you have to deal with a case where something goes wrong and U you have a desync now the other thing is of course Network effects right once we have the the efficient synchronous composability between chains we have the ability to build Network effects on one chain and use that on another and because ethereum already has massive Network effects that means space rollups can use that those Network effect X those liquidity users protocols Etc to U directly improve the the user and developer experience on their rollup now the other thing that we want of course all the time is faster things and cheaper things so on the faster side we're looking at good pre-confirmation that don't break things um and that are fulfilled 100% of the time frequent block proposing the more we touch ethereum the more we can compose with ethereum custom front ends so this is kind of changing the the unisoft style swap interface to something that actually you know represents what a pre-confirmation might look like so that's a green check mark um some confetti of some sort it doesn't seem like a lot technically but it's actually really important to improving the the user experience materially one thing is like getting rid of improval and then swaps we can get rid of that completely with pre confirmations uh another thing is of course cheaper things which is better execution so just writing your smart contracts better and U not making kind of the mistakes that we've made in the past when we were rushing to get code out aggregate everything there's a slide on that so I'll wait um and of course efficiency of of all shapes and sizes now we want to do all of these things and and both of these things I guess without sacrificing decentralization and censorship resistance liveness and sustainability so this is where you know you can get faster and cheaper if you go to all layer one you can do it on a centralized sequence or Layer Two what you can't do is faster and cheaper with decentralization sensorship resistance liveness and sustainability without some form of Bas sequencing is my my opinion so we know what we want it's not that complicated we want fast things we want cheap things and we want everything to feel like one chain for developers and for users getting there next Generation base sequencing this is the subject of these three companies up here this is kind of the the industry that we're in this rise spy labs and Tao gwith um and we're taking a very practical approach to unifying ethereum so we're trying to avoid the kind of discussions where we get into the Ivory Tower semantics um and actually build like tangible useful products that are good for users and developers so let's get into the specifics this is a intermediate talk so we're going to get a little bit technical here we're going to share literally everything so when it comes to um things we share like if the answer is no then you're probably lying we have deposits proposing blobs proving Network effects assets contract calls offchain infro security economies of scale etc etc etc so share deposits that's AG these are teams that are working to reduce cost of you know posting ZK proofs to main net by aggregating ZK proofs also has the excellent benefit that you can do interop between layer twos without ever touching layer one so you save a lot of costs with that as well shared proposing saves a lot of costs and enables Atomic composability again between base rollups if we can share blobs so if you're a base rollup and you're frequently checkpointing or posting to ethereum you're going to be purchasing a blob for da in most cases and blobs are pretty big and most Bas rollups do not use 100% of The Blob space in a single blob so there's a lot of empty space um and there's a whole bunch of teams spes kind of leading some of the research on this side to figure out how we can have multiple rollups use the same blob so we build some cryptoeconomic system to make everything a little more efficient and a lot cheaper for base Roll-Ups we also look at things like shared proving to to save cost shared network effects so using the layer one for Network effects instead of the layer twos um and we kind of logically centralize our network of X for a better user and developer experience and also of course shared assets so we want the canonical um ler of Last Resort for every asset to be on layer one um we want fungibility between these follow-ups that's something that we get with share deposits but it's a big priority in of itself contract calls of course so we want to do contract calls uh we want to batch those you know of infro that's pretty obvious security that makes sense and then of course economies of scale in everything we do so one of these kind of tools is pre-confirmation we have a over over here in number one we have a centralized sequencer that gives out pre- confirmations this is what every layer 2 does today except for Tao um and number two we have some proposed models for based pre- confirmations where the layer one proposer I mean like you know the guy running on a res Pi with a dialup is the sequencer and they're the one directly giving out pre-confirmation and directly interacting with the uh the people who are requesting pre-confirmation so obviously a pretty huge bandwidth and compute resources cost um and number three is the design being explored today which is some form of Delegation that's that purple line uh to a Gateway who actually gives out pre-confirmation to users and then they you know create a some Builder constraints some set of rules that Builders must build blocks to they comply to right so the the gayway is acting somewhat like a relay in this case but with a little bit more constraint around the builders so we still have an auction taking place um very similar to met boost and of course the Builder gives the block to our ethereum proposer who goes and gives that out to layer one now the other thing that we want to do is M retention um censorship resistance uh you know being included here and of course doing this all withe offs which turns out to be a really difficult problem uh but we have some excellent research from the ethereum layer one about execution tickets so we've kind of extended this to Layer Two on base rollups we have these things um that look very similar to execution tickets we distribute these in an auction and then you must burn a ticket to propose a layer two block uh although we do some of the the burning and um kind of registration early in the EPO and offchain optimistically so that there not an expensive thing to do onchain and of course with this we can throw in some no delay forced inclusion so every rollup today has a drift where you can deposit make a deposit transaction on layer one and then you might wait a few layer one blocks before that's included in Layer Two in a Bas wups case you you can pretty straightforwardly build no delay Force inclusion right into a uh a base rollup which gives you um you know excellent sens resistance equivalent to that of ethereum now one other important thing about next Generation based drops is checkpointing so whenever we tyo cause is proposing by the way but whenever we want to kind of put our sequence onto layer one we need to make a checkpoint here and this is little red flag on my diagrams but uh what we're actually doing is making um kind of a a point of as atomicity right a layer one evm transaction is atomic you can't revert part of it um and because of this you can do cross chain contract calls um based on that atomicity you can also checkpoint together that's this uh bottom left diagram where we share lots of checkpoints in one which saves a ton of costs uh the other thing we can do of course is other things while we checkpoint so we can use layer one liquidity during our checkpoint make this available on layer twos we can also um you know save the states of things on layer one we could change the state of a layer one contract to simulate um layer one execution but actually doing the execution on Layer Two cool so one important kind of piece of technology ology and probably the most important for the base rollup teams uh to compete on and be the first to is validity proving and fast validity proving and share deposits so there's a whole bunch of ZK teams working on this um cinct and Rio are two examples and the goal is to take fastf D proofs for a whole bunch of layer twos combine these aggregate these into one fast ZK proof and then put this onto ethereum in a share deposits Bridge contract that has funds for lots and lots of rollups now this has security kind of concerns you have to trust your ZK so we probably want multi-pro and it's also today really slow so you know the the biggest competitive advantage of Bas SCS in the future will be how fast is you're proving cool so this is kind of the the magic of Base sequencing that we've been working on at Spire to some extent of course and that is using a sensor positionist committee to sacrifice liveness for sensior resistance you can read more about this in our light paper um and then forward sequencing which is using the uh a sequence posted in um in a layer one slot to affect a future layer one sequence this enables you to do based sequencing and traditional shared sequencing at the same time which is obviously super powerful so if anyone from espressos in the room take a look at this cool so after we've done all this what do we get Harmony and more specifically deconstructing economies of scale to enable parallel Innovation while maintaining Network effects I've said this in all of my talks so far this week and it's so important right we don't want economies of scale to introduce centralization risks and we want to promote parallel Innovation as much as possible the other thing one of our important goals for user and developer side is a monolithic experience right not nothing monolithic in practice um but we want to have the ability for users and developers to feel like they're on one chain um just to to get that clearly better user experience and of course doing all this while establishing infinite expressivity giving builders the direct ability to customize whatever they want about the rollup including execution environments gas fees Etc infinite Gardens break walls next Generation based rups actually solve problems stay based oh here's a quote from heraclius Drake uh stay based for one man's liquidity is another man's liquidity thank you [Applause] okay thank you a lot uh so we have a few questions um just uh yeah so if you play cool so first question what can Tao do to become a Next Generation based trup uh will first of all upgrade their Bridge contract so it can share proposals and deposits and ZK proofs with other rollups and then add actually good pre-c confirmations um yeah they have a lot to do okay yeah second question as someone new to the space I find it hard to factor out the marketing how does your approach compared to Espresso sound similar at least that's a good question so espresso is building a shared sequencer um and they're building a form of what they call based espresso which is not actually base sequencing um they just call it that it's a marketing thing and the goal of espresso is to unify layer twos the goal of Next Generation based rups is to unify ethereum and layer twos so we're involving ethereum in the equation um but to get there you can do shared sequencing you can even involve the layer one sequencer um but that doesn't mean anything unless you do this checkpointing unless you do this Atomic synchronous composability so base espresso will not allow you to do cross chain contract calls okay thank you uh do you have any other questions uh I still have time okay sorry I could haven't figured out to log into that thing um so you mentioned that there is a no delay Force inclusion uh I'm kind of curious how that works with a chain reor uh where the chain State kind of change after like if you force like a deposit for example and a chain State changes with a reor right so this is something that every base rollup must do which is reorg with ethereum so in the case of an ethereum reorg then your base rollup must always rework um doesn't have to change anything but it will rework and this is something that is very hard to fix in protocol so we need an ethereum upgrade to reduce the risk of reorg or um decrease time to finality but there are a lot of outer protocol things we can do so if you get lots of proposers opted in in the look ahead you can just ask them not to reorg have them put up some collateral and you'll be able to reduce the risk of reorgs um you can also like paste over it on the user experience side you can have a a solver or somebody take a duration risk um the user never has to experience a reorg and then spy is also working on like an RPC that's custom kind of retrofit to deal with commonly common reworks one thing I would add is that Bas trops will rework as much as ethereum does now so I don't know how much you use ethereum but if you do a lot of D trading you still probably never experienced a reorg that has affected your um your life so I wouldn't expect that it's a massive user experience problem yeah we would have to rework okay thanks and then sorry just another question um I thought um cross chain contract cause was very interesting uh but I still can't imagine how it works like you know across let's say like a bunch of different chains that all operates at a different speed right so could you have maybe provide like some uh practical examples with like a bit more details yeah come talk to me afterwards because that's a you know a longer discussion um so all base rups have a batching time of 12 seconds and a block time of theoretically less right and this means that the kind of composability you do between them can only happen at batching time because that's the only time when you have atomicity and and synchronicity right and so we can compose every 12 seconds we can do a cross train contract call every 12 seconds um and then within that kind of boundary we uh we can't do any composability um that is enforced by cryptography right we could still do like intents or you other kinds of um interupt okay thank you please uh give a round of applause to our speaker he