Ethereum needs native L2 by koeppelmann | Devcon SEA

[Music] all right yeah so this will be um a controversial presentation um and I kind of expect uh to make uh well kind of everyone angry because uh I'm uh criticizing to some extent and the existing uh2 and to some extent the existing ethereum um core um road map but um I think it needs to be done so in this in this role uh I'm I'm founder of nosis but in in in this capacity I'm just speaking as someone who has built for over 10 years now on ethereum so we started building on ethereum when it was still a test net um deployed one of the first apps couple of weeks after ethereum was released and have uh continuously been builders in the space um so let me make a few comments on on on on the previous presentation I have the absolutely highest respect for for Jesse for um what Bas uh is doing what coinbase is doing in general but I do think uh the claim to say uh um we are bringing people to ethereum we bringing the next billion people to ethereum is wrong we bringing the next people billion people to base and that is a significant difference so just a few examples so Jesse was also talking about the 30% uh um fee that centralized platforms uh can charge from you of course referring to Apple that is something for example Bas can absolutely do as well so if you are app on base it's absolutely in their hands to control um how much fees you're paying for a transaction and of course they can start uh charging a 30% uh cut and this is typically something that doesn't happen early on that happens much later once a platform is established uh a lot of kind of lock in effects happen and it often even doesn't happen with the first generation so ultimately base is coinbase is a is a regular company and yeah there are great people I absolutely again have highest respect for for for Jesse and and Brian and but they are employees at the company and at some point they might um leave they might retire they might do something else and then another set of people comes and ultimately it's a company controlled by shareholders um there's a fiary duty to to maximize shareholder value so oh we have this asset base oh we can Spike up fees uh we can take more fees from from those applications and in a way it's even our duty to do it because we have to maximize uh value for our shareholders so my claim is bringing people to those L2 ecosystems is quite different from uh being on ethereum the claim of course is always rollups inherit the security uh of ethereum and we see labels like it's secured by ethereum you kind of get what you get uh that you would get from ethereum but the reality is absolutely none of those um that uh top l2s actually yeah kind of truly inherits the security of ethereum so if you look into the detail funds can be stolen here can funds can be stolen here funds can be stolen here funds can be stolen here funds can be stolen um and then on top of that now we are saying okay let's do chain abstraction so let's make it invisible to the user um on what chain they actually are but in a world where the underlying security is quite well different and opaque chain abstraction actually means in that case risk abstraction there was nice talk uh two two days ago so if you're abstracting the chain you're abstracting the risk and that's well that's uh that's dangerous so uh at some point funds might be lost and then you can no longer abstract the abstract the chain and you need to tell the user oh sorry um your funds are um lost but there are even more uh risks that even L2 beat does not cover uh yesterday there was um a great presentation by by James uh showing and again just referring to the um to the presentation showing that on any chain with a centralized sequencer that's again for example based arbitrum optimism if you have funds as a user on a money market like AA and uh compound the sequencer can indefinitely and at zero cost prevent you from accessing your funds from withdrawing your funds and and you might think yeah but you can Force include a transaction via L1 no it that uh won't help so roughly the idea is you create this withdraw request um and because the sequencer ultimately creat uh can control the exact transaction ordering they can alter the state of that money market so they can just for for a subc borrow uh too many funds out of that money market so that your request to to get funds out or your own funds out fails and then kind of directly after you repay so they pay zero interest because it's it's it's the same time it's in the same same time stamp so just kind of those kind of things uh yeah in my view again truly show if you if you uh build on one of those two uh systems you far from inheriting or getting um the security um that ethereum promises so what does ethereum security actually mean it means things it's a culture it's a process it's not it's it's much more than code so uh ethereum security means public core develop calls public discussions multiple client implementations rigorous testing highest level of of Promise um of backwards compatibility and kind of those things cannot be um derived or exported automatically to tol2 just because you um yeah kind of deploy some some code on ethereum that process and that culture cannot be derived of course there can be efforts to kind of promote that culture and to some extent ethereum is doing that and kind of talic is going around and say yeah if you don't do those things then I won't call you an L2 anymore so yeah that culture can we can try to promote this culture and L2 be is of course doing a great great job in in kind of trying to export that culture but again it cannot be derived um um with code so looking at another data point um again the claim uh assets are secured by ethereum but the reality is um only or it's it's it's the part that is actually of assets that is actually coming from ethereum is is getting lesser uh or the fraction is getting smaller and smaller so most assets again this is for example base but it's very similar on on uh on on other chains and on on smaller Chains It's even much worse most aets are actually not coming from ethereum so this is here the canonical um uh the oh yeah so so back here in time uh n the canonical Bridge was still the dominant so that that is kind of the one the assets that are actually kind of secured by ethereum now uh the dominant part are native assets so those are assets issued on the on the L2 or on on that other chain directly and they do not uh inherit there's not a concept of exiting um those to ethereum um and there are reasons for that so if you use the canonical Bridge uh and you currently and and you want to derive that security and you currently want to say let's say send a message from uh from arbitrum uh to optimism and and get a message back that mes that transfer took takes two weeks so kind of if you actually use economical Bridge it would take you two weeks to do that so A pigeon is faster in transporting a message for most parts of the world so yes no one is using that mechanism instead we're using uh new kind of bridges or external bridges that don't inherit the security um of ethereum so if most assets um are not natively bridged and sequencing is also not done by ethereum so as it's not bridged means not from ethereum not secured by ethereum sequencing not done by ethereum then the role of ethereum kind of um yeah uh is reduced to being kind of this checkpointing uh um yeah checkpointing um systems so here um again rollups could choose to uh yeah how how they how they uh build blocks and they can either choose to use their own sequencing um which means they can do super fast confirmations uh they capture their own me um and but the disadvantage is that they can only um not synchronously read uh into ethereum and that is a rational choice to do if you optimize for connectiveness to other chain or for um uh yeah for for treadly or you can go kind of what's called based um where you let ethereum do the block building and and the sequencing and here you are really optimizing for connectiveness to ethereum um but that's only rational if it's more important for you to be uh yeah as closely connected to the kind of economic zone that ethereum is but the reality is only 1% of um of of of value um is is is choosing um uh to do that so here is my my proposal for ethereum to uh to fix that and to address those issues and the proposal is that ethereum itself should develop and deploy uh ZK uh proven EVMS rollups and deploy 128 equal instances of that that are highly oper interoperable and that are truly kind of if you built on those you're truly built on ethereum so what would that mean being built by ethereum it would mean that don't even think about introducing multii that would be Unthinkable that ethereum deploys something and it's it has some some multis um as the upgrade mechanism it would mean that we at least have well we want to have multi client so we would have at least uh two independent implementations of let's say yeah the proof systems that's the most important part part here rigorous testing thousands of eyes that look at uh at the stuff and actually care about all the details that um are are are yeah often Pulled Under the rck um in uh so the idea is yes we can still have l2s wide range of designs within that we have based L2 so l2s that use ethereum for Block production and for well um and for yeah sequencing SL block production but even a subset of those would be native L2 so truly built by ethereum governed by ethereum and again we can here see uh some some things so um uh reads into L2 are synchronous here um and but of course also the economic perspective is important so uh so just it only in a native rollup essentially all value of that rollup is is captured by well ethereum itself now one question could be is is is that actually um sharding what what I'm I'm I'm proposing here no not exactly so sharding would be a design where you would have um in a way multiple L2 instances and that was kind of still the promise a couple of years ago that we would have 1,24 uh shards so essentially instances of ethereum L1 uh and and they would all be kind of live on the same layer what I'm proposing is is not that but instead but still something that comes tries to come as as close as possible have the L1 and then have those yeah 128 equivalent um l2s but still um in in this hierarchy and looking more at it um yeah some some uh things so so on an L1 you can do synchronous reads and writes and that's what we all love about about ethereum this uh composability the composability uh within this this larger system would still be pretty strong so from an L2 the idea is you can still do synchronous reads into ethereum so meaning you can execute some code on the L2 you can actually read a contract uh on L1 in that execution so let's say in price Oracle or or any any anything that lives on ethereum and you can kind of even yeah read a function continue with that result in the in the L2 process um and and do things so kind of on the L2 can do things dependent on the state of the L1 so it has immediate transparent access the other way around from the L1 you cannot read immediately the state of the l2s that is asynchronous reads but wres are actually synchronous so if you can very well do a transaction um on L1 that in the same transaction so atomically also creates a right into the L2 and affects the state you will just not in the context of L1 be able to um to immediately get kind of the result of that state change so again the read will be um asynchronous same uh in communication between l2s so you can synchronously make a transaction that affects the state of two l2s but again you cannot read from one L2 uh to the other so the general idea is stuff that everyone should be able to read um in kind of contentious State might live on on on the L1 and on all the l2s they can access the state synchronously um but there can be much more um non-contentious uh um state so all the things I have proposed so far would be possible today on ethereum without any any um kind of upgrades or changes uh to ethereum it would essentially just take the the um yeah kind of the ethereum community to decide um to build that of course if we do that we can actually make those l2s um even more powerful so few things or two things I'm proposing here that could can be done um if we make those L2 native we can and make them also native to the economics of ethereum so ethereum of course let's say issues eserve to reward um validators to participate in the consensus in the same way it could also kind of redirect or direct this issuance uh towards um proving the correctness um uh of all uh of all those um l2s another one is um that I'm suggesting uh that those l2s should have distinct name spaces so meaning an address should be clearly um attributable to to one L2 or to to one uh kind of in a way chain either L1 and L2 uh and in my view that is a big problem that right now you have an address and it can and the same address on on various chains can mean very different things so for example if you have a safe um then actually it it can the same address can exist on multiple chains but it can have completely different owners so what I'm suggesting here is that we shouldn't have this address uh collusion so each uh L2 um kind of uh uh yeah uses an additional salt to uh to to have its own address space and if we have that um we could we could allow sending a message from the L2 with its unique address into the L1 and actually have um have the message sender um uh on the transaction be that unique address that only exists um on on on the L1 so that allows you to do things like let's say you hold a token on L1 by an address or a contract that is only lives um on an L2 but yeah um some technical details so here's kind of the Spiral um I I expect or I see already happening uh um if we don't go this route so if we don't go this route it means um the economic zone that is ethereum and ethereum block building um becomes less relevant and therefore it already makes it less um less attractive to use ethereum as a block Builder or as a sequencer and again we already see this today 99 of the economic value of L2 choose not to be based and not to use ethereum as um a block builder at the same time more and more um Native or assets are not coming from ethereum and are not kind of really secured by ethereum either they are natively issued or they use external Bridges so for those the economic security of ethereum also matters less and less so in uh in total that means that um the relationship between rollups and ethereum really becomes weak weaker and weaker and yeah I I'm going so far and say it just becomes a meme so now so essentially in my view um that's the crossroad we are uh in front of with Native rollups we can kind of continue to have ethereum the most relevant or economic zone I would I would claim that should be the goal to be the most relevant economic zone even in the world in in in the world that's a place where prices are there's also the other perspective some believe ethereum should be a meme or is a meme um and they promote this idea of um yeah let's say iser is money and their perspective might be um if we do Native rollups we kind of damage the meme we damage our Collective um maybe religion or or kind of and because we we create tension within this uh uh nice family that we are that we all are so we disenfranchise those rollups so therefore the East meme gets weaker and all all that economic stuff doesn't really matter it's just about saying is's money and spreading eer and therefore it's better to uh uh to not do Native uh rollups all right so final uh final slide um for whom is it to decide uh what route we go absolutely for for all of us um that are here for the for everyone who is who is ethereum and we are all ethereum it's not just um let's say the core develop some core developers or or the ethereum foundation um essentially um it is um yeah ethereum anyone can essentially affect changes um uh to ethereum well thank you very much thank you so much and uh while we're still on that note we have some questions from everyone here and the first one is there areund 28 identical rollups which one should I deploy my dap on yeah um so um so I mean the first important part would be that right now you if you want to to to uh to deploy somewhere right now well you either use L1 ethereum or you have to choose one of those many ecosystems be it super chain or egg layer or or something and this would give you the opportunity to actually just say Okay I want to be um on on ethereum now um if you um uh if you um have to choose a particular one then again reads and writes within a particular um L2 are are synchronous so if there are other applications you want to regularly interact with then yes it would be wise to um to choose that one if if you're fine to say okay I just need um I just need to uh uh access I I I want this close access to the L1 um they are all equally close to the L1 then you really should just choose the rollup that is used the least um and and because that will be um um the cheapest why not implement this on nosis yeah I mean um maybe but but but I think really I mean the um well ethereum is is is is is the big thing so ethereum is that um uh has that credible chance to be this most important um um uh economic zone as as much as I would like nosis to become that um we we are not there we I mean ethereum is 100 times bigger and um and and again I mean um saying that again so there are all those ecosystems and NOS this is one of them that that that also tries to do uh valuable things and again I think that's totally valid I'm just saying I'm not saying if you come to nosis you come to ethereum no you come to nosis and we try to do valuable um things there and in the same way uh I would say other ecosystems again uh all the they are great ecosystems and they bring uh Great Value but you should understand then you are building on that ecosystem and not necessarily um on ethereum all right why 128 Why not start with four or 16 yeah I mean um the the idea is really to uh to make it clear that building um on ethereum or building here um is is a long-term viable and that um and and and and that of course addresses costs so I do think well I absolutely also want to share the vision to bring a billion people to uh on chain and um and and and for that we ethereum needs to be much more uh ambitious and I think 100x in in effect increase in effective block Space is really what what should be aimed for on a timeline for let's say two years I I I would uh say could be realistic for that um and not not a Forex or something like that all right the next one wasn't necessarily a question it was more a compliment they just agree completely and they're thanking you for stating the hard truth um but then we've got do you you consider this as rugging existing rollups yeah I mean uh to well rugging yeah um so I think I think existing rollups um well already have big time Advantage they alive um they are already built so that that will take at least yeah again I think let's say another two years so but but then yes if existing rollups don't will will not offer anything other than kind of just provide evm space then they will have a strong competition but I am uh I do absolutely think that all those uh all those yeah existing rollups are great people innov in Innovative people and I think they will be able to uh to kind of evolve and and um and create something or or essentially have additional Innovation on top and if they do that then yeah there can just be much more uh kind of Base uh basic evm block space as we know and love it since 10 years that's what I'm proposing here to just create much more of that and then there can be those l2s that are more Innovative that do additional things that do Cutting Edge maybe buil-in privacy may maybe some form of other sequencing that is is much faster there are all kind of ways um uh to to innovate so I think yeah again both both absolutely can uh can live all right unfortunately we are out of time thank you everyone for your questions and thank you once again Martin for your phenomenal talk