The Limits of L2s

Transcript

maybe I can already start talking um about what I'm going to tell you once my slides are up um so I have already been introduced as the NT uh L2 speaker and this is not true so I love attitudes I think l2s are a fantastic part of the solution of scaling ethereum um I just basically if you take home one message today I want it to be El chuza not enough to scale the ethereum verse and this is ah fantastic thank you so much fantastic so what I want to talk about today is the limitations in trade-offs that Crews come with so

um maybe let's look at the history and Genesis of l2s for a little bit um it became clear pretty early on that ethereum on its own was not gonna scale especially if as it was envisaged at the time and I still hope it's going to come true um is that ethereum is going to become the global settlement layer for the world so oops now it's going too fast no in the wrong direction ah fantastic so if you've been if you've been in this space for a while and you know that

we've talked about charting forever so back in the day sharding was actually um used to mean something different than what we think of shrouding today so back in the day um we were thinking of shards as these smart contract execution charts that could actually execute transactions on their own not the data availability shots that are currently on the roadmap with Dank sharding and I think the uh basically the the conclusion of this talk will be that this is kind of something that we kind of need to go back to to kind of go to that vision of kind of interconnected charts to kind of scale the ethereum ecosystem as a whole

so um Roll-Ups um became a huge part of kind of scaling ethereum pretty early on resources were poured into rollups you know to to a very large extent and there is one roadmap is now fully committed to scaling via layer twos since then we've actually made some pretty significant technological advances so we now have trustless ZK Bridges between different evm chains um and maybe had we had them back in the day or had we known that they would come way earlier than anyone would have thought possible

um maybe this would have you know uh this would have unfolded somewhat differently so this is talk about the limits of l2s and why we need more layer one block space Also okay so Layer Two is at this point they're securing billions of dollars but mostly their means so in a way um they are because they are unmutable from a single multisig and they have to be and I can tell you why just a little bit later in the talk um but basically just the fact that they are mutable from a single miter Sig drives very much home the point that other funds that are secured by those Protocols are inherently at risk of

compromised wallets so there's also uh the fraud proofs that are not operational and optimism at all and that are permissioned on arbitrum probably know this um shout out to the awesome guys at Edge who beat um so if you if you're interested in these kinds of details and how things actually work under the hood they are a fantastic resource I rely on them constantly good so a roll-up only architecture so what are the shortcomings originally Roll-Ups Were Meant To Be

liminal spaces so the idea was that you would have like say 100 transactions and you would kind of instead of settling them one by one on ethereum mainnet that you would kind of roll them together and then settle them on ethereum as a roll-up it's even I mean that's kind of where the name comes from right um so that works super well for some kinds of applications so things like um exchanges Futures exchanges that works super well because basically you trade back and forth back and forth back and forth and in the end you kind of you have a single number that you kind of settle to L1 this is actually fantastic instrument for scaling for many deaths doesn't work

for all depths it depends on whether the state of the DAP itself kind of bloats um or doesn't so for instance like an exchange where you kind of trade back and forth and then you settle like a final number um this is an example of where it works well there's also examples of where it doesn't work well so let's look at the limits to scalability so you've probably seen this on Twitter it's made the round quite a few times in the in the last months so if you looked at Bitcoin and lightning together famously it does not have sufficient bandwidth to even open a stage channel for every human at Birth

and close it you know you know after 80 years when they're presumably dead so that's not that's not um not that's not even kind of possible um as is now also basically which kind of very much um drives home um the limit of the scalability at the moment ethereum and layer 2 is on much better so maybe let's do some back of the envelope maths ethereum currently has on the order of uh 1.25 million transactions obviously depends on the kind of transaction transactions so this kind of varies a bit by day but there's nothing on the roadmap

um on the you know core Dev roadmap for the next say three years or so that will actually significantly significantly change that number so let's look at some examples so ens names you all have an ens name but let's assume 10 of the world would actually want to register an ens name for themselves that's 800 million names um and so if ethereum did nothing else nothing but actually register ens names and I'm not I'm not talking about renewal fees or updating content or anything or actually doing anything but ens it would still take two years to actually register all of these

second example there's currently 45 000 publicly listed companies in the world that you can trade um if basically these were to be settled on ethereum despite I mean they can be traded on their twos and so on um but if you you would probably have to have some settlement between Brokers or movements you know between the different attitudes um there would there would be less than 30 transfers per stock per day um and this is not talking about um uh that's not talking about uh I mean basically all of these are you know one or the other right so basically what you do nothing but kind of settle these

kinds of things also the fees on layer 2's they rise in lockstep with fees on layer one that's because the layer two's inherently they have to um they have to post transactions to layer one that's kind of how they're designed right so basically if L1 fees rise L2 fees rise lock step so um on layer Twos for a regular transaction currently the fees Spike above a dollar um regularly and that's unacceptable for many use cases this is going to get better by dank sharding eip4844 um it will probably get like 90 percent cheaper but that kind of that gives you

a factor of 10. basically if demand Rises as it has in the past factor of 10 is nothing so kind of we need more we need more okay so second Point um stock users so um if the cost of actually exiting your Layer Two um exceeds the value of the user's assets on that layer too the users I mean not technically but economically stuck right and even if everyone could afford to pay to kind of um to off-board of the layer tool back onto the layer one

um not everyone could do it at the same time it's kind of like trying to evacuate like a skyscaper skyscraper through um like a single fire exit door so um that kind of leads me to the next question what kind of applications actually lend themselves to being rolled up so there's many applications that don't actually now it's just going I think I just have to speed up so I can keep up with my slides now um so there's actually many applications that don't lens itself that don't lend themselves to being rolled up one of them for instance is circle is Ubi it's um this decentralized web of trust Ubi there's several hundred thousand

accounts on uh it's it runs on noses chain um there are several hundred thousand accounts and basically that and all the trust connections that's impossible to kind of compress and Export to layer one um another example is poapp um that all of you are probably familiar with and kind of meant by the dozens um also things like um social media apps so for instance lens um uh wouldn't lend itself to kind of being off-boarded to L1 or things like um your even your nft collection to be honest so now that actually leaves us at a point where for many many things it's actually

unfeasible to ever exit the layer tool so basically you you kind of you choose um to live your life on an L2 so it's become a permanent space it's no longer limited space um so there was this uh this tweet by anska basically roll-up Centric roadmap summarized you will live in an L2 you will consume the blobs and you will do nothing on L1 and you will be happy and I think that's currently the vision that um the way that it works on many Air shoes but I would argue that for that the cost of actually being on an attitude is too large um compared to

being on another L1 okay another Point centralized sequences sorry guys um So currently um all Layer Two Solutions currently have centralized sequences I don't know whether people kind of understand generally I how important that is to appreciate that means that exactly one entity is allowed to actually build blocks so their level of decentralization is literally zero the all layer tools kind of planned to get to you know a decentralized sequence eventually but right now it's not right

now it's literally one single entity building blocks so um this is I don't want to knock coinbase because basically they've been they've been battered enough for the last couple of months but and I think basically moving um kind of in the more decentralized realm with bass I think there was a great move for coinbase but I think it has to be said um that obviously coinbase controls the centralized sequences and it it is always kind of a party that can um I guarantee its own applications and its own order books priority over others

right so basically there's like the the central idea of kind of building things on blockchain is to to a certain extent um that everyone's the same and this is very much not um the case on uh tools at the moment we can also see this here so this is actually not an Edgewood this is ethereum mainnet and you can see the ofak compliant blocks over the last half year or so um and you can see that basically at some point they were like in excess of you know 85 percent or so um that were olfac compliant and effectively would Trend would censor

transactions just because um the uh the relayer that people used um the flashbots three layer um kind of was built that way so this has gone down over time because they have they have the repos were open sourced and a couple of people kind of launched their own relayers their own censor so there's um agnostic relay by Justin Drake and our own um sorry agnostic relay by us and uh and ultrasound relay by Justin Drake so basically it's gone down to like 27 again but even on ethereum kind of there are um the veil of centralization of decentralization can be thin so even and I'm pretty sure the flashcards people

they didn't set up they didn't want to sensor um any transactions it was their lawyer that told them dudes you you got to do this or you know you risk whatever and um so basically at the end of the day it kind of came down to the decision of that one lawyer and that kind of led to 85 percent of ethereum blocks being um in principle censored okay so let's talk about l1s so Bridges between l1s and l2s are fantastic this is why people love them right but if you appreciate that most of the

assets that live on l2s are natively issued on air Jews and are not bridged over from ethereum um this is this kind of this security Advantage Advantage goes away right so basically if you have an alternate layer one and You Mend assets there they are just as secure almost just as secure just by running your own note or kind of trusting that the services you use run their own notes so basically this is something that kind of gets repeated a lot that that other l1s are inherently less secure because they have a smaller validator set and so on but basically if you if there's a way for you to kind of

be a full node yourself this is not true so basically if you run your own four node um on another layer one um you nothing can be taken away from you so um on gnosis for instance the natively asset issued assets are circles that I talked about earlier monarium which is a euro stable kind usdc issues on many chains on many other l1s natively backed issues tokenized stocks and real world assets and there's only going to be more of those okay

the other thing is kind of the state of ethereum and how it's still evolving evolving and we don't want it to stop involving but basically as long as ethereum evolves um layer 2's will have to have upgrade rights right so they cannot be immutable because it's possible that things happen on layer one where basically that will break things on Layer Two right so one example of this um okay this is from Danny Ryan so basically he talked about the ossification of ethereum when basically he predicted that ethereum was going to change significantly still in the next five to ten years so um one example is Snapshot X which is

snapshot on stock net and so basically if ethereum goes to vehicle trees snapshot is on on starting net has gone a break so you need upgrade rights for start net to kind of make sure that you still have governance okay so what's the solution as I said layer choose we need them right this is it's not this is not an anti-air to talk it's an L2 plus talk so we need more so um what we kind of envisage at gnosis um is kind of to have um several many ethereum-like chains that

can that kind of have the same execution and consensus layer architecture as ethereum and can have trust minimized Bridges between the consensus layers so um the ZK Bridges and the ZK like time Bridges that's something that's new basically I think in the first proof of principle um was deployed last year between girly and gnosis and basically it's cryptographically secure bridging um obviously Bridges always come with different risks so um and if you look at Bridge hacks over the last say a year and a half or

something it's kind of in excess of four billion dollars that have been hacked obviously that is not a great State of Affairs so what we think needs to happen is we think there needs to be an into ethereum protocol we have built something to kind of um try to get to get there it's called Hashi Hashi means bridge in Japanese and what it does is it's in effect it's a bridge aggregator so basically it looks at the Block headers as relayed by different Bridges and it only lets you bridge if all blockheaders agree so basically as compared to using a bridge sink leaf and it adds a little bit of over a gas overhead

um but basically you have the additive security afforded to you by several Bridges who may have different mechanisms different implementations and so on so somewhat disparagingly kind of I would say it's a little bit like having a thief in a sieve in a sieve if you put enough sieves you know it into each other then you you end up with something that is super secure by additive security and in a way this is this would be equivalent um to what Cosmos has um in Cosmos it's called the IBC yeah fantastic good um so the IBC kind of trustlessly connects different um Cosmos chains and the incase ethereum

protocol can do the same for um ethereum chains um kind of fulfilling this smart charred Vision that we had like five or so years ago um so to summarize l2s Leverage The decentralization that the L1 ethereum already has and that's fantastic so L2 is a fantastic tools for many many things but we cannot stop investing in these energy in decentralization kind of producing incredibly neutral and resilient block space so that's kind of what we're doing um so if you look at the

decentralization stats here um ethereum currently has 574 000 validators and gnosis has 122 so that's kind of like 20 of what ethereum has how did we get there so we actually made it as easy as humanly possible to kind of become a validator or nurses chain so first of all the capital requirements are much lower so on so on ethereum you need 32 East which is you know in excess of fifty thousand dollars um and on nurses change you need one gno that's a hundred dollars you can run it on almost any hardware that you don't switch off

um we um uh we incentivize validators um in regions where we are underrepresented so there's gnosis VIP you can look it up at gnosisvip.com um so if um I I looked it up earlier and Montenegro was actually currently unclaimed so we have validators in 76 countries as opposed to ethereums 82 and our goal is to get to one validator per country so basically if you set up and reliably run a noses validator in Montenegro you can claim your reward for that so but I didn't actually want to go for what so

um but what I what I kind of want to drive home is basically currently we have 120 000 validators on nurses chain so basically for most intents and purposes that is plenty decentralization um so it's 2023 and you're deciding where to deploy at the app so there's definitely pros and cons of kind of all the different venues you can deploy right and I think questions you kind of um uh you have to ask yourself is how much does my offering suffer from platform risk so if that is a lot you kind of you don't want to go to something to a solution where you're at the mercy of a single entity or their

lawyer how much do you care about credible neutrality how important is moving between ethereum and your dap that's kind of the thing that really else who still have lack up so basically the bridges between l1s and l2s are more secure than the ones between different L1 so the interior protocol um I would argue that the interetherium protocol and basically find me anywhere I'd be here until Tuesday I'm super happy to talk about Hashi in the into ethereum protocol um this is not to talk about it but um yeah it's I I think kind of we've made we've moved

in leaps and bounds in terms of progress for Intel one bridging um do you rely on third party liquidity where's that liquidity do you rely on other doubts um and what's the transaction price point that you're that your adapt is economically viable at Okay so where to find out more we're Dao so basically come to our forum and kind of just participate in the discussion we have Dev docs you can look at the chain sets so as you can tell decentralization is something that is near and dear to my heart so kind of we have to make sure

that noses chain is you know is credibly neutral and and resilient and become a validator okay we also we're also hiring so if you are looking for a job please come seek us out and um just a very um quick uh as every year we are organizing a conference in Berlin during Berlin blockchain week this year called Defcon it's in September um and tip ticket sales are going live soon and speaker applications are already open cool thank you so much [Applause]