Justin Drake I Ethereum Sharing I Ehtereum Foundation I Pragma London 2024

[Music] Let's uh get started and please welcome our first Speaker Justin Drake on [Applause] stage thank you KK and uh good morning London uh I guess I'm based not too far in Cambridge so feel free to say hi if you're passing along so um today I'd like to talk about ethereum sharing and this is the context of chains using infrastructure or reding infrastructure that ethereum provides so you can have shared settlement you could have shared data availability and these are things that we're very familiar with in the context of rollups but it turns out that we can go much further we can have shared sequencing and shared execution and these are kind of mental unlocks that I hope to communicate uh today and by the way this image here is is this meme of the United chains of ethereum right the closer we are the more integrated we are as l2s as rollups uh the more Network effects we will have and the more uh we share the greater these Network effects will be so first of all I kind of want to recap um some of some of the the the existing sharing that's that's being done today and kind of give you an overview of the of the opportunity and then I want to dive into two specific types of sharing which is not being done today but it's an opportunity for the future specifically shared sequencing and shared execution so if you're a chain today um you basically have options to you whenever there's an ethereum module you have the option to share this module with other chains or to just go your own way and today I want to try and convince you that often times sharing is maybe the the right direction and like what are some of the the benefits of sharing uh ethereum infrastructure and it basically boils down to inheriting some of the the fundamental properties of ethereum layer one you you start inheriting security inheriting the the credible neutrality of ethereum and also Inc inheriting the liquidity and network effects and that's as opposed to introducing new security assumptions and folding down to some sort of weakest link introducing new brands that has maybe less Lindy and also just creating new boundaries and frictions that break down Network effects so what are the key um modules within ethereum the way that I think about it the four basic ones are settlement data data availability sequencing and execution and in some sense sequencing and execution of these these new ones now um for settlement we all familiar with this if you settle on ethereum you become an L2 like that's a very simple definition of what an L2 is if you choose to not settle on ethereum then you're an L1 you kind of separate from from ethereum and in terms of data availability you also have a choice either you consume ethereum data availability in which case you become a rollup or you choose to not consume etherum data availability in which case you become a validium and it turns out for these two other mod as are similar Concepts so for sequencing you can either be a based L2 or base rollup or you can choose to be an independent L2 or an independent um rollup and for execution um you know we just it introducing new terminology here basically is you can be a native rollup or native L2 or you can be a custom L2 en roll up I'm going to talk about that a little bit more now I also have some emojis here so Scrolls can be rolled up the way we have the scroll based with the sunglasses and and the crown and I think what we're going to move towards is this this world where we have all three so uh Native based rollups and I guess you could call them ultrasound or maybe um now there's all sorts of Downstream consequences to reusing some of these ethereum modules so if you use eum settlement then now suddenly you have your settlement contract on L1 and you can start composing with other contracts that are on L1 and so you can start using IF as money you can start using assets as collateral and so you start benefiting from the economic bandwidth you kind of economically plugged in and this is a great advantage that we all know and love you once you start also using uh data availability you start benefiting from you know finality and liveness so this this boils down to security properties and you start benefiting from the economic security of ethereum which by the way is on the order right nowadays of $150 billion doar um you know more than the order of magnitude greater than than Bitcoin and then if you choose to use ethereum sequencing as a module then you start benefiting from uh things like synchrony so um settlement here gives you asynchronous composability but once you also use etherum as a sequencer you you you have synchronous composability which is a a stronger form of composability which has fewer boundaries and I'll talk about that and that gives you shared liquidity for example so one of the applications is a Unis swap or matcha that works across all the rollups right or um sorry one inch for example so if you go on one in you know it will try and get liquidity from all the L1 dexes but it won't get liquidity from rollups and that's because there's like these strong asynchronous boundaries and that's the kind of boundaries that we can we can destroy with um with shared sequencing with dl1 and another really cool thing is this idea of shared pre-confirmation so right now if I have a transaction on arbitrum I get a pre-confirmation from the arbitrum sequencer if I send a transaction to optimism I get a a pre-confirmation from the optimism sequencer and these are separate counterparties that I have to deal with what if instead there was one counterparty um that could provide pre confirmations for all the execution environments for all the chains at the same time that would be a much cleaner and more unified user experience and then finally shared execution is going to be around having the exact same safety as etherum L1 the evm so every time you create a new rollup there might be a bug there might be you know attack vectors like governance that can change the rules and can rug users and what that means as a consequence is that we've lost fungibility right an ethon L1 is different to an ethon optimism which is different to an ethon on arbitrum because there's just different underlying security assumptions and so once you have shared execution you actually reestablish fungibility across the ecosystem and another really cool thing is that once you share all of these four you actually have shared forking meaning that when the L1 Forks you you also Fork so you have shared reorgs synchronous reorgs and you also have uh synchronous governance so the layer zero the social layer can change the rules of ethereum and all the rules change at the same time imagine having like different modules Each of which is independently reing Each of which is independently hard forking and upgrading it's kind of really messy and requires maintenance and this is one of the reasons why bridges are so brittle and and so dangerous and and and and and and get emptied on a regular basis and when I see this this diagram here I think of network effects and I just see network effects everywhere um and this is one of the reasons why I'm just so bullish on ethereum effectively becoming the settlement layer for the inter internet of value because there's such strong Network effects that there's an opportunity for one layer um to to grab the The Lion Share here and I believe that you know a FM has the capability from a scaling standpoint uh to to to to have room for everyone okay so now let's dig in specifically to shat sequencing this is a topic I've been thinking about a lot lately and this is this idea that different chains different l2s on ethereum can use the same sequencer what does it mean it means that at any given moment in time in any given slot there is a well- defined party who has Monopoly power power to order and sequence transactions for all the rollups that opt in at the same time and what does it mean to share sequencing with the L1 it means that it's the layer one proposer who's already doing the sequencing for the evm at layer one to also be doing the sequencing for these other chains now one of the powers of shared sequencing is that it unlocks synchronous compon composability so by default if you have asynchronous chains you have asynchronous composability and this is this is good for for some things but it's kind of junky and and and and and awkward and so really we want try to have things happen at the same time want synchronicity so one very basic form of synchronicity is synchronous inclusion where you have transaction on rollup a transaction on roll at B and they both get included on chain at the same time but we can do much better than that we can actually have the transactions be aware of the state roots of each other so you can have the post state route of execu of transaction a postate roote of transaction B kind of read each other's state from the different rollups but we can do even better than that we can basically have the various um rollups and l2s and execution environments send assets to each other immediately with with no friction all within the block so let me try and and visualize it so so today we have these silos we have the the layer one is its own Silo roller a is its own Silo roller B and basically There's an opportunity to do magic um there's an opportunity for parts of the block certain transactions to be so-called super transactions a super transaction is one that touches synchronously multiple execution environments now if we zoom in to this uh section here where the L1 and roll up a synchronously compose we have a super transaction made of three subtransaction so we have um the the the first yellow one here we have a green one here and we have a yellow one here and the important thing is that they're all happening at the same time synchronously and there's this notion of immediate transfer of assets so you can have a token immediately um be withdrawn from rollup a for example swapped on rollup B for cash and then the cash goes back to rollup a and all of this happens seamlessly just like it happens on layer one so on layer one I could have a token and I just send it to Unis Swap and then I can go through a or some other SM contract and all of this can happen synchronously so in some sense we're reintroducing what we've lost and in some sense we're trying to fix the fragmentation that the the the rollups have introduced in the in the short and medium term and I won't go into the the technical details but basically there's there's several ways to to do this and there's a a lot of recordings online that are now available and two technologies that can actually enable this are number one realtime proving so it's this idea that you can prove statements about the rollups in real time with ultra low latency and the second technique is um to use liquidity providers so you can imagine the shed sequencer to provide liquidity and Bridge uh any delays in the withdrawals that the rollups may have now another really big advantage of shed sequencing is this idea of shed pre-confirmation right you have a single counterparty and they give you a promise within 100 milliseconds that your your super transactions will go through so in some sense we're trying to recreate the user experience of Solana low latency uh user experience uh with with with with synchronous compos synchronous composability um on ethereum and the the the way that and this this is maybe like the biggest research unlock in the in in the last few months uh in in that space which is that we can do it with L1 proposes so we can have a subset of the L1 proposes put forward collateral and become sequencers and pre-confirmation so those would be like the um the L1 proposes that are in green here in the So-Cal look ahead the other uh proposes are just what we call includers and the idea is that if I want a pre-confirmation as a user I'm going to talk to the next pre-confirmation will be executed such and such and then we have the includer um basically include and settle on chain the transactions that have been sequenced as promised by the the sequencer again I don't want to go too much into the technical details but the long story short is that we can have the L1 proposers provide sequencing services for the the the rollups and this is kind of this somewhat lucky coincidence that we have this resource uh that can be used by the chains it's and it's somewhat similar to call data call data is a little bit of an accident that it's it could be used for data availability by the rabs it was not designed to be data available by the RO but it just so happens to be that it can be used as such and it's the same for the layer one proposes and in the last few months um there's been a lot of excitement for this this this notion of based sequencing um so uh we have Tao uh yesterday Aztec signaled that that might want become a base rollup a rise plans to launch as a base rollup and there's this mystery like top rollup one of the very top rollups in Productions today that also uh aims to become a base rollup and then we have all sorts of infrastructure Builders and and and researchers and and developers uh building stuff and yesterday we had a whole ethereum sequencing day with roughly 50 people uh where we we kind went through all the a lot of the technical details in order to to achieve this Vision okay now the final part of my talk is a little bit more futuristic but I think it it will happen this is the idea of shared execution so today if you want to build a rollup you have to build your own custom virtual machine even if you want to be evm equivalent you have to build your own circuits your own proof systems your own you know flaud proof game whatever it is and you'll have your own bugs your own governance and all of these things and there's this little kind of hidden box in thealex um graphic uh um of the of the road map which is called explore evm verification pre-compile and what that means is that once we've snar ified the evm at layer one meaning that for every eum block we have corresponding snot proofs that these blocks are valid and they can be verified very efficiently then there's a very easy Next Step which is to expose within the evm a pre-compile to verify evm execution so the pre-compile would take this be three inputs a pre-state route a postate route and a snar um that this state transition between the pre-state and the postate is valid and then the pre-compile would return true or false if this this s passes verification and this pre-compile basically allows you to build rollups n l2s that are equivalent to the evm not just simulated and like let's let's hope that they're equivalent but like actually by construction uh equivalent and here Chains would be reusing and sharing the execution of of ethereum and but in order to get there it's going to take several years because we need client diversity just like we have client diversity of the execution clients we have Gaff and ref and whatnot and you know because we're really paranoid in terms of security we also want to do formal ver verification and if any of you here are listening and are formal verification experts please come to me because we're considering setting up some sort of very large prize like tens of millions of dollars to actually prove formal verification of of EVMS and and ZK evm circuits now once we have this kind of magical evm within evm um pre-compile then the barrier to entry to creating a new chain is extremely low right because most of the heavy lifting technically to create a rollup today is around the verification game or the the verifier circuit we also have correctness by construction so we have no more bugs you can't as a developer of a rollup introduce a bug in the VM because it's literally by definition the evm itself and another really cool thing is that you have governance synchrony meaning that anytime the social layer chooses to change the rules of the evm and we have a hard Fork well this evm pre-compile within the evm will magically upgrade at the exact same moment as the fork happens and so now you don't need a separate governance token to maintain equivalence with uh dl1 and then the final really cool thing is that we can actually remove the gas limit so why do we have a gas limit because there's this denal of service Vector where we want the the people running the full nodes and and the validators to verify that blocks are valid within a reasonable period of time within a few seconds let's say but once we have SNS we can verify the vality of blocks in just a few milliseconds and it's constant time it doesn't depend on the amount of gas that was used and so now we can have we can remove the gas limit completely and basically be botoni by the fundamental limits of physics and and CPUs in terms of how much they can process and this is basically in some sense a return of the vision of execution sharding so previously we had this idea of having 64 shards or 1,24 shards each with a very homogeneous virtual machine but here we have programmable execution Shing meaning that anyone can trustless spin up A Shard and not only that but they they can customize it so it's it's much more in line uh with the programmable nature of of ethereum and not only is it programmable execution sharding but is like much more scalable execution sharding because we can remove this this gas limit now one question which is I guess worth asking is what infrastructure will the chains not share like what will be the custom izations that are unique to every single chain and I think there's there's a few of them worth highlighting I think each chain has an opportunity to create its own brand to have its own Integrations to have its own economy so for example base like has a very unique integration with coinbase and that's something that makes it very very very valuable um another thing is that the each rollup even though they might choose to share the execution with the L1 they still have a lot of room on the tokenomics they can have their own treasury their own token governance they can choose to do their own public goods funding like optimism does and they also have sovereignty over the value they create so all the execution fees and all the me can acre to these execution environments and this is this is one of the big departures uh with the old road map of execution Shing where all the fees would basically get burnt and go to E holders here um the idea of programmable execution sharding is where actually fees acrew to um to to the developers and the LTS and that's it thank you if I have time happy to take some [Applause] questions we'll uh we'll just do one or two questions if if somebody has time or if any questions sorry we have a couple minutes so U do you have any questions from the audience one hand in the back I'll have you kind of come up here and Shout cuz we want to make sure the question is on the on the mic so please uh please move forward and uh we'll give you the mic here it's on now sorry maybe while we wait um thanks a lot and you mentioned already it's taking time how long and what can we expect on the way on the way there so one of the really exciting things about shed sequencing specifically does it is that it doesn't require any hard fork and so it's just a matter of building the the tooling and infrastructure of chain and it can happen incrementally so you can have one rollup do it and then another one join in and so this is a very organic uh thing and in in six weeks uh Tao as I understand plans to launch the first base roll up and then a few weeks or months later after that they intend to turn on pre- confirmations so pre- confirmations does require a bit more work um and what I'm hopeful will happen is that Tao will be will pick a really good uh pre-confirmation um you know piece of infrastructure that others can just copy and then some of these other chains maybe Aztec maybe rise maybe some other chains can once they see the benefits of synchronous composability kind of join the party one of the interesting things is around trying to measure the the size of the network effects and one of them is is basically called melt cf's law which says that the strength of the network effect is quadratic in the the size of the network now let's look let's try and measure the size of the network using tvls a proxy the L1 set has $600 billion of tvl and you know the largest uh sequencing environment is arbitron's um centralized sequencer which has I know 15 or 20 bill 20 billion so there's more than an order of magnitude difference between the two and so the you'd expect the size of the the the difference in in network effects to be you know at least two ERS of magnitude and I think we have this very cool and graceful plan now which is or like at least possibility for assets when they migrate from L1 to L2 they stay within this shed sequencing Zone and they don't destroy the network effects of AUM because the the current path where we on where each rollup has its own independent uh sequencer is kind of chomping at the network effects and destroying it here while creating new ones but the new ones that are being created are so much smaller than the one that we started with mamy okay question is there a plan for a shared Constitution so when you share the L1 you know this's this idea of sh forking and part of it is that you're actually inheriting the social layer and the social layer can you know just change the rules as they see fit and in some sense Constitutions are for humans and and so in some sense yes you are sharing this ethereum culture of when you choose to to Cho to to build uh on ethereum whether or not we'll have a formalized constitution um I don't know about that but uh maybe hi justtin this what uh yeah um so uh I guess my question is uh so I see the upside of the shared sequencing and the collaboration um but it would feel like a shared sequence also maybe creates centralization risks like between the protocols and you know attack vectors so I guess my question is like in the event of that are there fullbacks that you know the protocol is going to have to like remove these risks right so one of the the common questions about Str sequencing is are we not moving to a world where we have a super Builder that's basically building blocks for all the rollups at the same time and I think there's several answers to that answer number one is that we want to make sure that even hyper centralized Builders can't do harm on ethereum so we want don't want them to be able to censor we don't want them to be able to cause liveness or safety failures we don't want them to be able to front run and sandwich and it turns out we have technology to make sure that even the most evil uh you know superb Builders can't do harm so for example for censorship we have inclusion lists and encrypted me pools um and so I think it is important to be pushing these Technologies in parallel to pushing the meme of Shar sequencing another possible answer is that we have extremely high centralization of the building today even for a single virtual machine which can only process 20 transactions a second we have three Builders uh Beaver Titan aryn that build 90% of the blocks and in some sense centralization is inevitable and so we might as well kind of accept it and the best thing we can do is be prepared for it and then there's a final answer to this which is that even if we have separate sequences for every single execution environment we should expect layers on top to be built uh which are basically collusion infrastructure for the proposes and sequences of these execution environments and the reason is that if there is a non-negligible kind of value to be had from synchrony versus asynchrony then now there's more of an opportunity to extract me and extract fees and so you could imagine for example collusion infrastructure like suav kind of being built on top and creating marketplaces for the sequencers to say hey if you want to buy up my sequencing slot at a premium I'll just sell it to you and and in this world we actually end up with more complexity and more centralization because the barrier to entry is higher precisely because we have more complexity to collude and so if we just accept the most simple solution to unlock synchronous composability which is just to have a single shed sequencer at every given slot in some sense that is the smallest barrier to entry that you could expect for the builders thank you