Ethereum's L2 Limitations: Need for More! | ETHDenver 2023

whatever now we are ready so I will be talking about um the limits of L2 so the talk will be um yeah will kind of challenge a pretty dominant assumption uh that's yeah dominant in the ethereum ecosystem right now and the idea is well you have ethereum and then the way to scale is to have um yeah a bunch of l2s and I will make up this presentation can actually take the mic yeah so I will make the argument that l2s are not enough and that we instead need need more than that oh no oh yeah okay so this talk will will yeah discuss um discuss the trade-offs of of l2s and therefore also to some to some degree criticize them but again my message is not that there's something wrong with that juice it's just that they do come with with trade-offs and they and those trade-offs should be acknowledged so for some use case we would argue um they are actually not the right um solution so let's first discuss um how we are how we came here so right now very obviously ethereum doesn't scale so kind of the dominant solution right now is um yeah is is to have l2s um in fact the uh the reason to do l2s was that um other Chains would always have the the problem that you need to Bridge and the bridging technology to other chains um yeah introduce risks and that's what l2s are trying to solve but there were significant improvements made for um for Bridges and we now have yeah zero knowledge zero knowledge based trust minimized Bridges so that somewhat uh will change is changing the equation um but yeah we'll we'll come to those conclusions later so first there is a very very obvious criticism of the current state of Altus and that is that at this point while the work is real the research is real but the current implementation of l2s yeah is what I would call mostly memes so the l2s that exist right now do actually not bring the security guarantees they are supposed to bring in fact um they are upgradable by emojisix so essentially you are trusting that multi-sig and you're not trusting um yeah kind of the actual um mechanism so there's this great website L2 beat that really goes into into the detail of the risk and kind of just to take pick the two biggest ones optimism is live but it doesn't even have uh the fraud proofs yet um so the sequencer can just yeah kind of submit any malicious State and there's nothing you can do about it and even up arbitrom um they have fraud proofs but first the fraud proofs can only be executed by um by yeah kind of permissioned basically there's a permissioning system only you if you're allowed to submit fraud proofs so well that's already one part but then there's also a modisex that can do arbitrary upgrades and actually this modise is doing transactions every few weeks so kind of every few weeks they could immediately just a theory wipe out uh the two billion um that are in the bridge but that's not what I want to talk about I want to talk about kind of let's assume all those things are fixed let's assume l2s are in their actual desired final State let's discuss those trade-offs so the original idea of um l2s was the idea of temporary spaces so you are on the L1 you have your asset on the L1 you make a deposit into the L2 then you do a bunch of transactions um in the L2 and eventually you come back and settle on on the L1 and and that's kind of the picture it's liminal spaces so spaces you don't really stay in you kind of just you pass through but eventually um you come back and that was that was the original kind of thing um yeah people said and kind of also that's where the name roll up comes from So You you do a bunch of transactions and kind of you roll them up and you can eventually settle back um on L1 so let's let's look into that so first of all um there's absolutely nothing wrong with this concept and it does increase uh scalability but there are absolutely limits to the scalability um that you can get through that and I think one one example that makes this very very obvious uh is is to look at Bitcoin and lightning because it's essentially the same promise so you kind of deposit into your light Link channel then you do kind of a lot of lightning transactions but eventually you um withdraw and some people well naive people on Twitter claim lightning therefore solves all scalability problems and if everyone would just be used lightning then uh no problem with scalability but if you do very very simple maths you immediately realize that that's completely because if you kind of just just think about it so let's assume everyone on the planet uh eight eight billion people would want to use lightling then kind of if even if you make the complete uh ridiculous assumption that they only have to do one protection but they're born and kind of to open up a channel and then 80 years later when they die they close the channel which is of course a complete absurd absurd assumption but even that could not be handled by Bitcoin so I guess simple argument is if we are talking about this idea that that l2s are liminal spaces so kind of um then the effects of scalability you'll get are still limited by by ZL1 chain and if the L1 chain is simply um too small to um yeah kind of to to facilitate this entry and exit um yeah again you have limits so discuss the thing with with a bit with Bitcoin let's let's look at ethereum so those are the numbers of transactions that ethereum can do um and as you see since a few years it's essentially blocks are full um and essentially ethereum is capable of doing 1.25 million uh transactions per day um and at least on for the next three to four years there's nothing on the ethereum roadmap that will will increase it I mean maybe we can a little bit increase the gas limit but fundamentally that is that is what we have and and that and that one won't change so kind of oops wrong direction so um I mean this is just just to give you some perspective uh if if if if we want to say ethereum is a settlement layer something like ens should live on ethereum then even a basic assumption like 10 of the world would want to have an ens name just to register kind of those 800 million names that would clock ethereum completely for two years so for two years you couldn't do anything else so kind of just just again to give you a perspective um if things should eventually be settled on ethereum then still that's that's a big bottleneck uh the other example would be again the idea ethereum should become the global settlement layer there are currently 45 000 stocks or kind of publicly listed companies in the world if we if we think ethereum should be the settlement layer eventually those stocks will be tokenized and kind of settle on the chain and again if you make the Assumption trading trading of those stocks this may be happening on L2 but at least the settlement like let's say for for large Brokers or even kind of just uh to kind of move from 1l2 kind of settle on L1 and move to another L2 then even kind of here you would just have 30 30 transfers per stock that are actually doable on ethereum so again kind of does it there are clear limits um with with this approach and and again I mean we don't want to do we we want to use ethereum for much more things than settled stocks so um yeah again just to give a perspective um this um those those yeah scaling limits of course are uh eventually um yeah kind of expressed in fees if if there is not enough or if there's not enough space simply the fees will rise and the current state is the fees on an L2 on ethereum are still proportional to the fees on L1 so right now most of the time fees on l2s are in the range of I don't know 15 50 cents but if these once in a while Spike on ethereum they also Spike or on l2s and they can easily go even above a dollar and there are of course plenty of use cases um where where a dollar interdiction fees is absolutely not acceptable and there's one use case we are working on um so we are actually creating a credit card that is also a hardware wallet so every time you do a credit card protection you are actually authorizing um and you're actually doing a blockchain transaction and for such a use case where you want to do your I don't know three dollar coffee payment with the card you absolutely need to have a subset transaction fees and again l2s on ethereum can currently and will also not uh in the future um offer that for it 4844 will provide improvements but still we won't get in the sub send region [Music] um this this concept that if you if you um that you exit to the L1 brings another problem that um if the value you have on the L2 is small is smaller than kind of the the exit fee or you kind of the transaction that um that it would cost to exit to L1 you're kind of you're kind of stuck in the L2 so it's kind of just economically not um not feasible um to exit the next argument is not every application can actually be rolled up in such way so again we're talking about you deposit you do a bunch of transactions and eventually you withdraw but there are applications um that simply create a very rich state so one one project I'm involved with is kind of circle CBI but don't need to know too much about this project but essentially it creates this very very rich and this is actually a screenshot of the of the web of trust so it creates a very rich State people kind of trust each other make connections or I mean even even some projects where you create a lot of nfts those are kind of applications where it's not really possible or where you don't really have those efficiency gains because um if you generate those thousand nfts that then and you would eventually access then them to ethereum there's actually not much you're saving you if compared to issuing directly or uh uh on on ethereum because again remember the advantage was that you can do a lot of transactions but eventually you just need to settle one with one value but again that's not possible in every application so let's move on with the argument the conversation has a little bit changed uh of around l2's and again initially it was this thing deposit eventually withdraw but what if you never have to leave to what if the twos are not liminal spaces but are permanent spaces so Roll-Ups is permanent spaces and I think this is also kind of pretty much the general yeah general idea or general Vibe uh in the um in the ethereum uh ecosystem now yeah um level to consume the blobs you will owe nothing on nl1 and you will be happy so let's look into that so most of the or essentially all of the l2s that exist have a proposed centralized sequencer so what is the centroid sequence so it's essentially one entity that um that can produce blocks and yes they have they are constrained by the L2 so there are mechanisms in place that prevent them to uh to produce invalid uh invalid um blocks so they cannot just straight steal your money but they still have a lot of power so what they easily can do is say your transaction I don't include I I censor this transaction or your transaction I take a little bit a higher fee or this application I trade I treat um different so yeah very very obvious example is um is this uh um base uh project and again don't get me wrong I really think uh bass is a great improvement over the status quo I absolutely do think that uh users having their assets on base is much better than users having their assets on coinbase but you should not mistake this L2 for credible neutral platform so if they wanted to they can easily um they can easily uh kind of yeah for example as as described in this tweet here uh combined attacks um with with the centralized order book and they can do so because they control the sequencer and well then good luck competing with with another decks on this uh on this platform so again it's I think it's absolutely a well a good but a good platform to have but it's certainly not the neutral platform um and yeah well thanks to you guys so they're um there is famously even on ethereum even on ethereum which is um well it's supposed to be uh very decentralized even on ethereum we saw the issue that at some point up to 80 of the of the um blocks were censored and it's very very obvious that this is very real risk um that a sequencer um might also face yeah or will well will be forced to censor I mean we should also be clear here so people don't do that or most of the time people don't do that because they think they they wanted to do it but because their lawyers are essentially there's pressure on them or they basically have to um so I think there's a very very real risk that um yeah sequencers will have to follow sanctions or or even go so far that they require users to kvc and say kind of we only process um process transactions from kyc users and then again yeah in theory you can exit or kind of go through the through the L1 but well at significant cost and then that's not really the the platform you want you want to be on or well again it depends I mean I think but I think one should just be aware of this uh of this risk yeah another argument that's a little bit more complex is um the idea of where is an asset natively issued and and the argument is if we if we now think about it so initially the idea was we exit eventually to the other one but now as we are permanently on the L2 so then the idea is if the if the asset is issued natively on the Chain why is this linked to the L2 so important after all so to make it more concrete uh on it on a chain um I mean usually the argument for l2s is the risk of the bridge but for assets that are native to the chain you don't have this bridge risk and kind of just to make it concrete um on on those chain for example we have various assets that are natively issued on the Chain so again if the circles ubia project I talked about earlier but we also have a stable coin that's actually considered e-money but whatever so it's it's native to the chain um and they're a very big argument Falls away because the big argument is 40 L2 is kind of to say you inherit the security of the validator set um of ethereum but for an asset that is native to the chain even if on uh yeah on the previous slide even if 80 or whatever whatever number of validators on let's say noses chain would um would kind of say or what would make a false malicious State transition your full note would still uh not accept that state so as long as let's say the the monarium guys kind of those who are issuing the token they run a full node and you run a full node um then no matter what the major or if they might even if the majority of the validators would go malicious those full nodes would not accept that malicious state so so the argument is an asset asset natively issued to this the the chain and actually running your own full node is the highest level of security uh you you can have okay next argument is um Altus um or we discussed initially that ltus currently has this upgrade problem that they still have those upgrade rides and they will probably have to keep that for a long time and that is because ethereum itself is not yet stable and in fact it's a quote from from Danny Ryan um yeah ethereum researcher who said he himself is not expecting ethereum to stabilize or kind of to ossify uh in the next five to ten years so quite likely or I mean it very much seems like ethereum will continue to change and that also poses a challenge um to Altus and I explain why so for example there is there has been this L2 that does a snapshot Vote or it kind of you might be familiar with snapshot kind of with the voting and they they do this voting trustlessly on L2 can if you do the voting on The L2 and eventually you settle back to the L1 however what they do is they do Miracle proofs in the in the L2 um but guess what on the ethereum roadmap probably in one to two years and there's the idea to switch ethereum kind of the way ethereum clients work from from Merkle trees to Vertical trees essentially it will then make it easier to create those proofs so that's great but it would break it would break that solution so so it means if you are if you're building an L2 you will likely have to have upgrade mechanisms for the case that that ethereum itself um changes so so again it kind of well is in conflict with the idea that the l2s are as trustless as as hoped so finally um after discussing those shortcomings what is um what is the proposed solution the proposed solution is yes we we do need L2 so it's not about saying um it's not about saying we don't need them but we also need and well surprise surprise I'm I'm from those chain so of course we see uh our chain is as part of the solution so we do think we need more L1 block space and more decentralized uh one block space that actually offers um this um yeah kind of credible neutrality and uh I think we are quite proud that that gnosis is very deeply uh integrated in ethereum and we see noses as part of the what we call ethereum verse in fact you can actually use a standard standard ethereum client never mind Aragon guess and you can just run it with a network flag and and it will run run noses kind of that's how how deep noses is actually integrated into into ethereum so what what makes what makes all of that um uh much more feasible is the fact that we now can actually have very secure um Bridges between those States so what we were able to do in there currently at least four or five teams who have well already working implementations maybe not yet in in production but very close to it um it is yeah zero knowledge uh zero knowledge based um yeah proof of consensus so what this bridge does it it runs a full light client of ethereum hypnosis chain so in-nosis chain in the evm we can actually prove um the consensus and prove that this is a block that was signed by the majority of of the validators so we have a very secure way to know um yeah kind of to get essentially to read the state from ethereum and we can do the absolute same thing the other way around so on on ethereum we can run again in a zero knowledge proof we can run this light client and get and validate the current state from ethereum so that's absolutely a game changer and that um essentially combines the ideas of of Cosmos with the idea of ethereum so in the cosmos I mean Cosmos that was always part of the solution or part of the idea that you have those chains that are connected via what they call IBC inter blockchain uh communication and it was exactly that that idea that chains can run a line client or verify consensus of other chains and that has been much more challenging for the ethereum um client but yeah luckily there are so many smart people who figured it out and now essentially we can have the same same interoperability for what we call um the ethereum verse so chains that use kind of this execution execution environments of the evm and um and the consensus the beacon chain so the conclusion of the talk is um l2's can Leverage The decentralization so we have this great decentralization uh decentralized chain ethereum and we can leverage it with their juice to create uh essentially more of that block space but that leverage has limits and therefore we cannot stop to invest in decentralization and we need to produce more of this what we call credible neutral block space and that's essentially um yeah the mission the mission of ethereum so we actually just launched this dashboard to to kind of measure our decentralization um we already have a hundred thousand validators in I think yeah well 60 countries uh our goal is actually to go further to um to have validator ideally at least in every country at least one well data in every country of this world um and yeah so here are a few of the the trade-offs so again say it again not necessarily the twos are false but um you should you should consider it it's it's an option and it's a choice and kind of depending on what things you value um maybe maybe yeah a chain like noses chain uh which yeah we are explicit about we are not in L2 um is the better option thank you