Decentralizing Sequencers Wait it’s all PBS Always have been - Hasu

foreign [Applause] [Music] PBS yeah uh always has been so in this talk um I will come into roll ups are in the process of decentralizing on their sequencer so every roll up um has a decentralization roadmap and all of these roadmaps have some plans for decentralizing the sequencer however this uh the way that sequencing works and they are too it actually combines uh what is the equivalent of a layer 1 proposal and layer one block builder in the same role and this creates a variety of problems unless we start to address it by separating these two roles on Layer Two just like we did on layer 1. in addition to that even if we do PBS on layer 2 there are novel challenges involved around privacy cross-domain mov and um lower latency than we are used to on layer one and so we will get into these in the talk as well um and we would see PBS is essential but it is in itself not enough we also need to decentralize the beta Royal itself so decentralizing roll up sequences what is a sequencer um they are basically four steps involved so as a user you send your transactions to a layer 2 sequencer who orders them according to some policy and then gives the user some receipt to the user so italic was talking about pre-confirmations that is step step number three here and then the sequencer sends the ordered batch to the D a layer where it basically it becomes finalized from the perspective of the layer 2. so this particular diagram is taken from stagnet but the above is how it works in pretty much every um every Layer Two so there might be some small differences for example um in stock net a the sequence size also the prover so they have even more responsibility but all of these can be stripped out and would be stripped out over time so why is decentralizing layer 2 sequence as important well because as vitalik was alluding to because Roll-Ups are becoming more and more important and more and more um volume and liquidity is moving over to this new execution layers and I don't think we have even started to imagine the end state of um of how blockchains would scale as we're kind of seeing with these different Roll-Ups roll-up providers were now not even just focused on building their own rollups no they are focused on they have all pivoted to kind of Roll-Ups as a service and um it's I think it's quite possible that we will see a future where um spinning up a new blockchain is is kind of like spinning up a new smart contract today and so we will have many Roll-Ups in these roll ups all will have um sequences so we left many sequences and today all of these sequences are centralized and centralized sequences um can sense abuser they can go offline they can charge arbitrary prices they are regulatory choke point and they can see the user transaction before it gets confirmed and they can do various things with this transaction that is hard to attribute and so wild one of safety's great guarantees so we can get always a transaction right through the layer one and the sequencer cannot force an invalid State transition and so on then there's still a lot of problems um with them and that's why we need to decentralize them there are various proposals for how to decentralized sequences um one is the layer one or quote unquote other chain sequenced approach so that includes the base sequencing that vitalik was talking about there are other approaches for example about shared sequencing so that's when one blockchain um sequences many other blockchains so shout out to Espresso for example we have put forward this Vision um you have the optimism style sequencer slash block auction um where the right to propose the next block or an experience of blocks is auctioned off every now and then um and then uh what stock stock at is planning random selection from a proof of stake set so in that case you have if basically consensus mechanism on top of your of your roll up and a proposer is selected randomly and and they get to propose the block this is very similar to how it works in in the layer one and then finally you can have various committee based Solutions and the most common one would probably be uh first come first serve this is where Committee just looks uh what transaction on the different notes are first and then come to some kind of consensus on that and um yes we were saying already in the uh in the summary in the layer 2 right now uh what we understand as the term sequence are actually combines um the equivalent of a layer 1 proposal Builder especially in the leader election mechanism so that is The Proposal Part and then the ordering mechanism that is the building part and so if we go back to these different um proposals that have been put forward so they actually have different leader election mechanisms different ordering mechanisms but what I would like you to take away from this um from this slide is only that the innovation that is inherent to these proposals is almost entirely in the leader election mechanism so nobody has any credible plans about the ordering mechanism right so they are only focused on their leader election mechanism today and that is a problem because if you do layer twos without proposal Builder separation then what you do then you have the two roads bonded like they used to be bundled in ethereum layer one before Mev was a big deal but now Mev is a big deal and what would happen is you recreate the Dynamics of pre-pbs days from ethereum layer 1 on these roll ups and what you get you get priority gas auctions again you get front running you get Fair transactions and a clocked up P2P layer from the search or strategies who are being forced to compete in inefficient Mev auctions for um for The Med and because the octane is less efficient you will also get a lower proposal Revenue that leads to a thread of vertical integration between searches and block Builders so all of these problems that we talked about two years ago on ethereum boom they are back now earlier to making their comeback and so that's why I'll actually say when we talk about decentralizing sequencer to date sequencer as a service I think it would be more accurate to actually talk about proposal as a service because that's what it really is right it's we are only looking at the proposal today so we need to start separating these two roads on layer 2 just like we did on layer 1. here's a quick refresher for you what PBS actually does so the effect that PB has on ethereum is it Shields the proposer role from the centralizing effects of Mev by making the most valuable block available to all proposers equally and you don't whether it's a small Solo validator or it's a big mistaking cool like night or coinbase they all make the same from Med and that is a huge huge achievement it also unlocks competition on features from block builders that don't require any protocol changes to ethereum and this has created more privacy for bidders more expressivity for Betas and TI would particularly highlight the introduction of the bundle by flashbots that decoupled the position of the transaction from the price because in the public mempool you are only able to express basically these two Dimensions through a single unit which which was the gas price um the removal of negative externalities from the chain that were the result of these uh these bad Searcher strategies and um maximizing revenue for the seller and so we can see this map one to one to the problems that we laid out earlier um PBS solves them all which is amazing so we need PBS and explicit Meb options to maximize protocol revenue and minimize negative externalities on this layer 2 networks and their users but PBS on Air 2 also faces Noble challenges that were not necessarily here when we designed PBS on on layer one so the big one the arguably the biggest one is privacy so all of these centralized sequences have gotten the user very addicted to the idea of quote-unquote easy privacy right and um now I think it's a very hard Choice like do you remove privacy when you decentralize the Builder or do you try to decentralize it I think both paths are very very hard and it's it's not clear that either of them um is necessarily better or easier than the other and um and I I think this this kind of speaks to um the difficulty of Roll-Ups just taking these shortcuts through a centralized sequencer in providing the users better ux and now it's it's going to be very hard to either move away from that ux or um decentralize it so how do we recreate privacy in a decentralized way so um one is maybe you should have centralized blocked us but at least these centralized block Builders are competing with each other in a in a sort of competitive market right but the downside here is that these Builders all have different they all have their own own unique form of trust and so as a user always a blockchain you need to decide which Builder do you trust which build do you not trust and this really risks enshrining some Builders over others and it also creates a big incentive for vertical integration between different supply chain participants because you're always more willing to trust yourself than you are willing to trust someone else and so this option I would argue is very centralizing second one is again the formation committee based Solutions like threshold encryption um or first come first serve these models work together and then option three which is having one homomorphic so I mentioned the second one because this is basically a way of privacy and easy mode if you will because it's much easier to go but um let's stand under the idea of an encrypted mempool or a any system really where um audio back okay any any system really um where different parties share the same type of privacy where if you trust the the Privacy Zone then you can also trust everyone within it and so um for example what this would mean you send an encrypted transaction to a mempool and you know it's going to be kept private no matter who interacts with it going forward and so this is what we're working on with Suave but it's also super hard then you have cross domain and EV just when you think you're out they pull you back in um even if you have competition between these centralized Builders you still have the problem of cross domain Mev the Builder Market is already naturally very centralizing um due to economies of scale um but layer 2 is just because of the existence they make the problem even worse because the more fragmentation the more domains there are the more fragmented liquidity and and trading volume is the more cross domain there would be and this creates a strong centralizing pressure on the builderall through demand for Atomic transactions and then finally there is latency this I would put in the same category as as privacy so roll ups um they all wanted to have or at least at other German they wanted to have um based sequencing originally but then um it turned out that users liked faster confirmation so much that they have had to switch to a centralized sequencer with faster pre-conformations um and this is lower latency this is a big friction with decentralization we need geographical decentralization for our neutrality and decentralization and other things but for that we need our systems to be insensitive to latency um and if you lower the block time too much then what you get is you encourage vertical integration geographical integration between proposals and block Builders and you discourage participation from anyone who is not able to vertically integrate who is not able to co-locate in that same geographical area and so that's why PBS is essential but it's not enough we also need to look into decentralizing the Builder role itself so that's what we're trying to do with Suave um there are other proposals as well so but the idea is building the best Mev auction in making it decentralized because that's that's what allows us to do the heavy lifting on ordering for these other chains removing all elements of trust and flashbots and Searchers and block Builders from the system making it private so that participants have the ability to trust the auction and Trust others and don't have this need to vertically integrate with others in the supply chain and making it scale a bit because by connecting many different domains we allow for the expression of cross domain preferences as well so John charbone our host uh has this great um diagram here that actually shows the roll-up transaction supply chain in the future possibly and and it shows that the sequencer really becomes the proposer right and you can have proposed us that that basically sequence many chains but you still need the block beta and because you you have these centralizing effects on the Block that you also need the block Builder to be decentralized so in the roll up end game we're gonna need all three we're gonna need the decentralized sequencer we're gonna need the decentralized Builder and we're going to need the private member uh otherwise it's it's not going to work out and dear roll ups please make Mev a more central element of your decentralization road maps remember sequencer equals proposal plus beta don't ignore the last part we need to decentralize both so come reach reach out to us to learn more about Suave and come to our Forum collective.facebook.

net um that's it from me um thank you very much and um yeah I'm happy to to take any questions