MEVconomics for Modular Blockchain Stacks - John Adler

foreign [Applause] [Music] we will now be moving on to our next one we're gonna have John Adler come on and he's going to be talking about me economics for modular blockchain stats okay so let's get started uh so today I guess I'll be talking about me V economics uh for modular blockchain stocks uh and I'm going to start with what appears to be blockchain 101 uh but if you bear with me it'll become obvious why we kind of build intuitions by construction starting from the beginner uh so what are some components of any blockchain protocol not just a modular blockchain one it has a leader selection algorithm and these are some of these things uh will you know will you'll have heard of them from previous tops uh especially the ones that we're going to do dive in today uh so this is basically another one that tells you who is permitted to actually produce new blocks there's no blockchain that can't applicant of new blocks added to it isn't particularly useful uh then tying into this you need a simple resistance mechanism uh in any kind of decentralized blockchain you need some decentralized permissionless blockchain you need some mechanism to prevent someone from just spamming a bunch of potential candidate leaders uh and that's going to be the simple resistance mechanism so things like proof of work group stake Etc then you have a block validity function or a state transition function uh effectively the rules around the validity of the execution of a block independently of who God to produce it and then finally uh to wrap all of these up together you have a fortress rule which uh which allows you well by you I mean the nodes the distinguish between two otherwise valid chains so two chains that have the correct leader that has sufficient civil resistance of like that amounts whatever you want to call it and the Blaster all valid uh you need some way of distinguishing between these two otherwise completely valid Chinese all right and we're going to focus primarily on the first and the fourth fourth one of these uh because things like block validity yes the specifics of the blood for the function determine what Meb may be extractable for instance Bitcoin may have less extractional mov than for instance that you're in complete blockchain like ethereum but given that just a few days ago we saw York on bitcoin because of an achievement maybe that's that's not actually the case anymore uh the leader selection and the Fortress ruler definitely things are coming deployed uh with MEP so uh this is the only slide that will contain theorems and crawleries uh so please bear with me but the first theorem is that in any decentralized blockchain it's required that leader cannot hold the chain on their own uh statistically of course you know if we're talking about proof of work uh this one person gets really really lucky they can just produce every block forever uh you know sensor normal chain Etc uh why why do we want this to be the case well because of one party acting alone can completely and permanently hold the chain uh then they effectively have the ability to destroy it and if they can destroy it then they can they it's isomorphic to having control over it uh as as a coach from Dylan would say uh now uh corollary to this is that if you're given some some uh if a actor is given access to sufficient civilizer sufficient of like the Civil resistance mechanism amount uh they should be able to become a leader for this one block in some finite time again statistically uh and as long as this is true then you know the system is permissionless so decentralized permissionless is all it's all very good the general way this is done is I mean those leaderless protocols like Avalanche but you know here we're not going to into two too many specifics the general way that's done is that if the sine Theta doesn't act within some finite time bound then another entity becomes a leader okay so why is this important uh the reason it's important is that the leader can capture Mev either directly or indirectly uh through things like Arbitrage funds running sandwich attacks Etc now generally it's hard to manipulate who is going to be the leader because the leader is tied to a civil resistance mechanism for instance proof of work you can't just like you know show up and say hey I'm the leader uh you kind of have to do a whole bunch of work similarly introvers like you know various previous state protocols like tender Mint or gastro whatever have pre-assigned theaters you can't really manipulate the leader than any meaningful or cheap way but there's all the things you can manipulate which is going to be the Fortress rule which works in tandem with the leader selection algorithm and it could be easier to manipulate the four Choice rule than it does to the later selection algorithm and manipulation of this could lead to things like Time Bandit attacks Etc then of course you know once once you've rear with the chain then you're also the leader of this block so then you can also engage in front running etc etc we're within a single block or a sequence of our sequence of blocks uh now these may not be that much of an issue in an L1 context but we'll soon see why there's such a concern and why there's so much so much more flexible in a modular context so what is the modular world uh we're moving towards a multi-d chain and specifically modular world where you have shared data layers that provide a very high amount of data bandwidth such as Celestia or these two I guess ethereum nowadays and then a number of execution layers or roll up Sovereign scroll Austria just to name a few there's there's canvas of these already in development Banning already deployed Etc mouth uh in this multi-chain world or rather if we're really never really being generalist multi-domain world uh a lot of the concern is around cross domain or horizontal as I like to call it Mev where you have MAV that comes from Arbitrage and other opportunities within two different execution layers within two different Roll-Ups or two adding domains really uh but and this is obviously you know a very important problem Etc but it's not going to be the topic of today's talk or to I didn't talk I guess it's 15 minutes we're going to be talking about and exploring what I would call a vertical nav in the module of the blockchain stack which is the med relationship between the roll up and the data layer and this could of course extend to multiple Roll-Ups on top of the roll up Etc et cetera but this is like a vertical relationship instead of a horizontal relationship uh and now you might be looking at this and saying well there shouldn't be any relationship between these two because you know there's two execution layers trades are being done on one that might create Arbitrage opportunities on the other you know you can atomically execute transactions on each Etc et cetera uh but if you have an execution layer and a data layer the data layer itself doesn't have any Arbitrage opportunities because it doesn't execute anything so like how why would there be a relationship between these two and this is why this is an interesting problem because it's not cross it's not cross-domain mov uh in the same way that this horizontal MEP is it's very different uh and for this we're going to keep in the back of our minds the tongue-in-cheek first law of crypto dynamics.med cannot be added or removed it can only be moved from one layer to another uh or put another way and maybe it can't be you know destroyed it can only be changed from moved around from one layer to another this is of course not actually like a formula it's just a tongue-in-cheek observation uh when dealing with vertical Meb specifically not cross-domain I mean so uh when it comes to building a execution layer building a roll up on top of a data layer you need to have some mechanism just again on blockchain to produce new blocks needle leader selection algorithm right uh but you would also like to leverage the data layer for security because if you don't leverage it for security then you're kind of losing half of what you get out of the modular blockchain right you get you can get the ordering and you also get availability uh be nice if you can you know use use the blockchain also for ordering uh because if you only use it for availability now you have like these synchronization issues between you know one block change consensus and another blockchain consensus you would like ultimately all consensus to be done ultimately only on the base data layer and if everything else used some mechanism of merge consensus instead of their own completely independent consensus protocols because then let's say user completely independent consensus protocol you'd have synchronization issues so what do you want out of a later selection algorithm for one of these scroll UPS on top of the data layer well you would like to decentralize the sequencers the views of one sequencer you know the censorship concerns there's potential for mbb Monopoly and Meb extraction regulatory concerns Etc also decentralizing sequencers is a very convenient and easy way to introduce a token right just have a token some stake weighted mechanism round robin you know leader selection just like tenurement has uh there's a great way of introducing that uh what else you want do you want strong liveness guarantees you don't want it so that a single sequencer going down for instance means that the Rope goes down forever uh you'd also like to minimize trust assumptions here uh you don't want the trust that a single party is operating or you know small number of parties are operating correctly for the chain to operate correctly uh and also you want to minimize waste of the data layers block space and waste can happen if there's some sort of contention Etc where two transactions uh show up and one of them becomes useless and it just wastes the data layers from block space you you like to avoid that uh you what you don't want though is you don't want the mov to bleed down to the data layer and I gave a talk at I think that some maybe it may be so much back in but back a while quite a few months ago it's linked in the last slide of this so if you're if you have access to the slides it's uh the Youtube link on the last slide here uh that kind of Dives deeper into the soil dive too too deep but the intuition is that depending on the leader selection algorithm you can have Mev bleeding down to the base layer and why do you not want that because uh now you have a situation where you know they have a bunch of any B bleeding down to the data layer you have all of this Mev to be captured that instead of being isolated to the individual Roll-Ups is all going down to the data layer and all that mab to be extracted means there's some centralization pressure on the validators of the data layer to actually extract them and we would prefer that made be isolated as much as possible to the individual execution layers that create the Meb now there's a problem here and this is somewhat of a I don't want to say open resource problem but rather still an open challenge uh that we would definitely if you're an Mev researcher or et cetera if you're interested in uh modular blockchains uh please absolutely reach out to us uh so we can you know brainstorm on on this problem guarantee is that Roll-Ups don't necessarily care and aren't necessarily incentive aligned with the best interest of the data layer they have their own interests their own economic interest Etc uh and they may just do something bad for for the data there we can't exactly guarantee that they won't do okay uh so let's maybe go over a few uh examples of potential leader selection algorithms at slash four Choice rules that you could have for one of these Roll-Ups on top of a data layer uh one of them is first come first serve uh so the intuition here is basically a bunch of people just post blobs that are that have a particular namespace uh or that have a pretty you know let's say this blob is for this particular role basically when that's what I mean by namespace a bunch of people who post blobs completely permissionless permissionless lady the first person who posts The Blob to the Celestia blog for the namespace is the one that is decided this selected to be the leader exposed facto uh what are some issues here uh it means that the leaders ultimately decided by an auction on the base layer uh an auction says as we know been to a left-hand context uh while you're either at the trust Auctioneer or if you know the auctioneer is permissionless uh then the options can be manipulated but I guess options can be manipulated either case it's just a trust assumption rather than a permissionless assumption uh options can be manipulated so that's not great so this has the issue that it introduces uh PJ uh for who gets to be the leader and therefore bleeds the Mev that comes from that auction down to the data layer on top of that it also wastes a bunch of space uh low space because all these other transactions that aren't the first still go on chain and they're just not do anything for the application so they're useless like they pay fees yes but it means now more block space is being used without providing value to end users uh another example uh tender man style leader rotation this one is fairly straightforward is basically you take just the leader selection protocol tendermint not like the whole aquarium and the voting and rounds and all the P2P stuff just the later selection protocol which is like I don't know 20 lines of go code it's very it's very straightforward uh and then based only on the state within the roller so you don't use anything about the ordering it's left here or inside of that or the data layer rather more generally you use only the state in the roll-up and the Fortress rule is Trivial because well tender minutes work for you now those issues here which is that as a safety rather than leibnized performing protocol this kind of scheme makes it a bit more challenging to have good liveness because as if you use tender mint on an L1 you can increment rounds very quickly but if you're on a roll-up potentially you have to wait one or more data layer blocks uh in order to say okay this lead this the leader for the slot didn't actually post their block therefore we will you know up increment around and rotate the leader permissionlessly uh another one is highest Priority First this was a proposal that was floating around uh just a few days ago it would require a protocol change to Celestia or the data layer it's basically the intuition is introducing a field a priority field that can be validated by the block producers the data layer block producers uh and that serves to order the blobs instead of just a few use for instance or the position in the block uh now the issue is that it actually degenerates to being isomorphic to first come first serve uh except the auction instead of being done on chain is done out of band and opaque which may be even worse uh last example before I close it off uh his base Roll-Ups our base roll-up space are cringed or cringe uh so the intuition here is space Roll-Ups are based on uh work that I did uh minimal viable merge consensus from 2019 so like basically four years ago almost uh combined with Mev boost or some PBS time then the MVP boost is used to prevent waste of course you introduce additional trust assumptions since we know Meb boost isn't completely trustless uh so this is maybe a side grade rather than a strict upgrade over the work that I uh published four years ago uh now I'm sure you prevent waste into the trust assumptions what are the issues you still haven't mitigated the use of an auction so select the leader so Mev still lays down okay uh I was asked by Tina to add a slide on how this ties into me Mev economics uh someone will have to come after this and tell me specifically and like concretely and completely but MAV economics is but I guess this ties into Mev and the economics of Mev bleeding across a vertical blockchain stock uh a lot of the work ideas thoughts proposals that I listed here or didn't come for me they actually came from members of the Celestia team and others I'll list some of them here non-exhaustively so there's Conor rutual Evan Callum and Gabriel uh they worked very hard on these kind of things if you have access to those lies there's link to their socials uh if you click on the bubbles here so please follow these people on Twitter GitHub or whatever they're very bright and they're the ones who did a lot of the work on a lot of the ideas and proposals here and to be honor to do further reading as a tradition uh in the blockchain space there's lots of prior art so I posted a bunch of links here with a variety of proposals intuitions counter points Etc okay uh that's about it sorry for being like a minute over