'MEV Scaling Ethereum Panel' - Vitalik Buterin, Justin Drake, Phil Daian & others

guys thanks eddie bye i don't really have a preference i'm [Music] happy to start the second part i guess sure let's start the second part and then have a panel actually people in the chat are saying panel panel so let's do the panel all right so um we have three panelists vitalik phil and myself hello hello um so i guess i was i was asked to prepare a couple of questions for the panel um so maybe i'll just go through those quickly and then we can kick off the discussion so um you know the the topic is mev and scaling and i guess my first question is kind of qualitative and my second question is quantitative so the qualitative question is is optimal muv extraction in the context of scaling a centralization force right because we have these block proposers in order to do optimal mev extraction they need to be aware of all shards or blockchains or rollups so they basically need to be a super super node um with lots and lots of computational power they potentially need a you know a team of analysts a team of phds whatever um and if you zoom out i i know how to kind of decentralize everything except this block proposing um and this is kind of one of the the last unknown so that's my first question second question is is quantitative and it's like what what will happen to the fee market when we when we scale like for example one of the questions i'm interested in in the context of of ultrasound money is do base fees completely disappear right when we have lots and lots of scaling um or you know will there be some sort of dip but then it will go back up um or you know what are other considerations around scaling is it is it possible that maybe because we have lots of scaling lots of shots there's going to be many many more fees and so this is even going to make the the centralization problem even worse um in the sense that not only do we have these these centralized actors but they're going to be incentivized to be super super large because of all the economic activity like if ethereum becomes like the you know the the financial hub for for the whole internet and the whole world that's kind of scary um so anyone want to uh talk take question one or question two um i think question two is easy base fees are never going to zero because there is always demand for um use v all sorts of uses of the blockchain if prices go low enough like you know at some point someone in some minor is going to spin up a service where the blockchains just get used as a store to back up people's cat pictures so like we don't know what that floor is going to be but no it's there right um but do you expect we'll be in this in this deflationary mode after for example sharding i would think on that if anything like the fees would be more they might be less individually but i think i think at least anecdotally we've seen every time capacity gets increased like there's a dip for a little while as right level of activity is now like too little but then like pretty quickly there's like a return to the mean at some point and like on net i think there's there's usually more profit for miners after each one of these cycles so i would at least personally expect the same thing to happen but it's hard to say i guess right right and this is definitely what we've experienced kind of historically um like the the total amount of fees just goes up up only um yeah as for the first question i think it is interesting and it's like super concerning i agree that mev is a centralization force i think there's a question of like how much of a centralization force it is and like how much of a advantage decentralizing is there going to be and like how much of that can we mitigate through either incentivizing like explicit decentralization or minimizing the centralization incentive uh like in the current form it's very strong of a centralization incentive i think um i think so for example like let's say without a network like flashbots which has like this searcher minor separation it's a huge centralization incentive because you have the miners that each have their own infrastructure and like unless you can reach the barrier to pay that engineering cost you won't be competitive and that's like what we were seeing with pools like ethermine kind of having their own mev infrastructure um i think adding in this more decentralized infrastructure where anyone can participate in either receiving profit or sending bundles is like a step towards making it more decentralized if that infrastructure is decentralized which today it's pretty much not uh but let's say it is right then it's that's a little bit more decentralized but there's still i think somewhat of an advantage from participating at more layers of the stack like as much as we try to flatten this reward function into being linear on the validation power where like with 20 you have 20 profit there still probably will be some economy of scale to size related around like your ability to build network infrastructure or have business partnerships with entities or whatever it might be like i still expect that curve to be a little bit above the linear curve like i expect size to continue being an advantage so i think the only thing we can do is like try to squash that down as much as possible uh but i still see it as being there and being a huge issue but i i'd be curious to hear what vitalik thinks of this also yeah and i guess i've i generally view these um mev efforts as being about like if centralization has to exist like concentrating it in a particular type of actor so that all the other types of actors can still be basically just like plug into that and continue to have economies of scale that are as so linear as uh as they were before i mean it's obviously an open question uh the extents to which that'll succeed and then the other open question is if like even if centralization has some particular layer happens like to what extent it does that turn into a risk for like actual underlying goals so like censorship resistance for example right like if there is say like one um you know like professional trading company of some kind that starts producing eighty percent of all uh uh block orderings then [Music] what like does that what practical ability would that company have to like say stop stop me from sending my transactions um onto the ethereum watching if they if if it decides that my turns out uh that like it doesn't want those transactions included for whatever reason like would i still be able to get my transactions included by paying an extra fee and then the the extra fees of everyone who gets this political kind of add up to be enough to uh over uh to let some other um aggregator get more get more reward will there be some alternative mechanism where i'll be able to get included with a one swat delay will there be something else i don't know so so here's a related question is like how do we what are the goals of ethereum like with regards to decentralization i know this is a very broad question but i'm sure you've thought about it because like to take that that previous line of thinking to the logical extreme like there's like eos and algorand on the other end and like clearly we as a community want to be more decentralized than that but like how do we set a target and like benchmark progress towards that and like design for that goal or how how do we even start to think about quantifying that or or promoting it yeah i think like one thing that we definitely needs to get better at is like actually kind of expressing our thinking about the ultimate goals behind decentralization right so like you know are we concerned about els because um you know they don't do the um the right rituals and and pray to the right gods or are we concerned about like eos for very practical reasons and my concern about the centralized platforms is definitely for very practical reasons right like if you take either els or even one of the uh recent eos like platforms that have managed to get themselves into a position of higher respectability than els like the question is if the elites of that community get together and decide to force through some protocol change that the community finds unacceptable so you know this could be some kind of censorship this could be like forking in some kind of um issue and stuff to like say themselves and claiming that they deserve it because they're the core developers like something um you know would be like preferential treatment for like kyc participants or like something you know whatever right would they be able to do this even if the community was just like really strongly opposes this and in these centralized platforms i think the answer is yes right because there isn't a culture of running validating nodes and there isn't a culture of that many people running nodes at all and so you know if the elites get together they can schedule a hard fork and they can even schedule like six hours in advance and then even um by the then six hours later um the users by defaults detect nothing and then maybe 12 hours later users start detecting things and by then like the hard forks already happened and the users don't really have a way of like quickly making a user-activated soft fork the users don't really have a good way of coordinating a rebellion and so like they've just basically successfully pushed it through and like what are you gonna do are you gonna revert like billions of dollars of economic activity in the last six hours and uh try to coordinate all of the um block explorers and pools that disagree before there's even more activity like you can't right whereas on bitcoin or ethereum like that that sort of thing would very easily fail right because like first of all the number of elites you have to coordinate as larger you know you have more client developers more core developers second there's all these users running nodes and then there's all these users that are actually verifying everything and those users would have to be convinced to download the software it would definitely take longer than six hours just because of sheer numbers and so like your ability to pull off that kind of like forced protocol rule change in ethereum or bitcoin is significantly lower right so the question i guess would or the one way of looking at the problem would be like would a single dominant mev extractor like change this calculus and like make it easier to pull off this kind of social attack now like clearly they would the mvv extractor would not be able to force through transactions that are invalid but you know they would be able to censor right and so like the practical question is just like you know like exactly how much ability to censor what they actually have and like you know what is that something that like users would be able to find some way around quickly enough so i guess what i'm hearing is that the the fundamental role of decentralization uh in in your view at least in this perspective is to to maintain this power balance between the privileged actors in the system and everyone else such that everyone else can check the power of these actors and and stop them from pushing through unpopular changes unilaterally right i mean i think that's a definitely like one a way of describing a large part of what the the value of decentralization is there to do probably not a hundred percent um i think like there's other reasons still like you said and wants decentralization too but it's definitely a significant piece one one of the the things that we've tried to do in the design of e2 is minimize the power of these privileged actors as much as we can and i guess maybe one way to tackle mev is to basically constrain the blog proposals as much as we can and i'm just thinking out loud here but maybe there's an opportunity to somehow um constrain the the strategies that they have for block creation so they they have to commit ahead of time saying this is going to be my search algorithm um and i'm going to adhere to it and if i don't adhere to it i i i can't create blog so i will get you know penalized um and then maybe there could be some sort of um yearn style you know open sourcing of strategies so anyone can submit strategies um these strategies kind of get run maybe in a distributed fashion and then the most like the winning strategies um a portion of the the income goes to those who wrote the strategies um so and then maybe there could be some sort of social pressure to push for at least the the centralized entities like the the pools to to adhere to basically at constraining themselves artificially uh in the context of block production so i think i think like that broad marketplace is kind of what flashbots is trying to build in a way but i think there's still like a lot of the variability and a lot of the power of these entities comes also from their infrastructure so like uh the network infrastructure like how do you police whether they heard a message or not or whose message they heard first i still think that is like kind of the fundamental source of like why they need to provide this input into the system uh and i guess there's a question there of like how much of this p2p networking attribution side is like do does eth kind of view as core infrastructure versus like uh you know kind of designed to be changed and customized um and and if so what are the desired properties of that layer right so i guess you know it's kind of the classic separation of data availability and execution i mean we can completely remove the uh the unfairness in execution but that there would be you know potential unfairness in the data availability and i guess one thing we could potentially do is look at these off-chain data availability committees and that that might be a way to go from purely centralizers in this one single actor to some sort of committee which might be more acceptable more palatable long term yeah so i think basically that's the chain link fair sequencing protocol is trying to bootstrap that from a committee i guess the question is just like what inherent trade-offs are there and like where does it fall in the l1 l2 dap responsibility funnel distinction like because because obviously as you go down that the design needs to be more general and work for a broader class of people like l1 needs to be the trade-off space that's most optimal for a huge number of people and then like l2 can be more specialized and the dap can be like super specialized so where where do we fit that in i guess is a good open question uh do you think l1 is the the best place oh i'd say definitely not l1 and the reason is in order to do that data availability and sequencing you need to have a really good internet connection because you're you know you're basically connected to all these mempools and whatnot um so it would have to be semi-centralized but i guess the good news is at the very least we can get this federated security so there's kind of a lower bound on the on the security we could get because when we when we you know select a committee at random at layer one we're assuming that these are raspberry pi's running on the home connection um so they can't really do like the high bandwidth data availability yeah interesting so there there is going to be like a natural specialization vector there i guess as as you said it comes from both the data availability issue and i think also the like network asynchrony aspect of like we want people all over the world to participate and you know we need this this network asynchrony is like a fundamental part of how we're going to coordinate them so yeah right okay so if we have a bit of time not sure we do tina interrupt me if we don't but like my my last question i guess is kind of a meta question is what are things we you're most worried about like what are maybe some of the known unknowns or maybe unknowns unknowns for which you have some intuitions what what are these scaling things that you're worried about in the context of every view i would love vitalik to take this one first because i think um and i think i think that's like like cross chain mev extraction issue that you mentioned is definitely one of them um right yeah um aside from that like like we all just like will demand for low block times leads to centralization issues like pretty much no matter how we satisfy that demand even if we satisfy it with uh fancy layer 2 schemes um and then also there's just like issue centralization risks on the like staking liquidity side like basically would um staking in ways that open up centralization vectors just like have fundamentally better capital efficiency properties by enough to encourage people to make stake in pools that are i guess custom well i want to say like custodial but like the reality is there is going to be a spectrum and like the question is like where on the spectrum what um would they be like would they what would it be most capital efficient to stake and falls where basically the custody properties of that if become such that they open up opportunities for attackers to like grab up and attack and control a large portion of the stake power but that's like less of an ndv issue and more of a d5 capital efficiency issue and what go ahead phil no good i mean one thing that just came to mind is the question whether mev is objective or subjective in the sense that like we're kind of assuming that mvp is an objective thing and we can minimize it um and but actually maybe like this some privileged actor not in the blockchain world but in the real world so let's imagine that you're a government you know you're privileged acted there and so you can do things that others can't do and um maybe we're gonna have like perfect equality on the blockchain but because these other actors have privileges we're gonna have this um subjectivity of mmv like one transaction might be worth zero to most people but to you know a government it could be worth a ton of money yeah i think that's super interesting like thinking about the full space of manipulation people will do because you see it in the financial markets where like people will use some weird obscure power they have and like also do some trade like over there and like these two things line up really nicely for them and they make a bunch of money and like something happens and people don't know so yeah i expect there's going to be a lot of shenanigans um in terms of things i'm most worried about i think it's the same as it's always been with mev which is that like goldman sachs starts the mev division and like they're just fantastically good at it and like uh everyone buys the license to like have goldman sachs tell them what order to mine transactions in uh because that's the best way and there's like not a network there it's just like very top-down uh you know so basically eth turning into robin hood in an order flow sense uh that's that's kind of what i worry about uh and also how this interacts with like large exchanges in pos uh like you know binance and coinbase and and how that also plays with the the correlated slashing uh and the correlated failure incentives like are those incentives kind of enough to prevent centralization is the meme enough of their existence like will they kick in at the right time and kind of have the right effects i think those are all open questions i have on scaling