Battle for Scale: Sharding vs. Rollups

Transcript

Uh, so Justin should be over there. Perfect. So maybe I'm waiting for the jingle, but I think you can you can already enter the ring and get ready. Take your mic. Uh, to the audience, don't forget that we have votes. So you can already scan the QR code so that you are completely ready to vote. Maybe I will just be in the middle. Yeah. So I will just be in the middle. Maybe you can be a little bit more on the side like the others. And then it's perfect. Well, uh, waiting for the jingle. Um, ladies and gentlemen, are you ready?

So, you you feel ready already. I can I can see you're ready. In the room is ready, but been ready since 2018. Perfect. Perfect. Perfect. So, the topic of today will be shing versus wabs. So we are going to make it entertaining I hope. So on my left we have Justin from the foundation uh champion of the consensus of the new beam chain uh who can say us a lot of stuff about charting just joking. And on the other side we have Ilia. So I think from Neil I think you are going to take the shing point. I don't know why. I mean just a thought.

Uh but maybe to start uh we can say that well shing has been quite the topic in the beginning of of Ethereum 2. I mean when the merge was announced uh sharding was supposed to be a big thing. Uh it was not really present in the road map or at least present in a different form. So my first question to you to both of you would be to well maybe define what challenging is today uh where we are how did it evolve and what are your opinion on that and maybe well what are doing better or well what is your opinion on all this? So let's start with let's start with you. Sure. Yeah. So you're right actually the Ephre FM 2.0 was all about sharding and this is what kind of got me into to research in 2017

and um I think one of the the big takeaways from the last you know more than half decade is that um there's a better way to do sharding and we kind of gave it a new name. Uh there's a lot of similarities but one of the big differences uh is around synchrony. So back in 2018 sharding was asynchronous sharding and what we've realized is that it's possible to do it in in this better way which is synchronous and uh what we've done concretely is what's called dank sharding as a first step uh which is synchronous data availability sharding.

Um and it turns out that you can also do synchronous execution sharding thanks to uh zero knowledge proofs. And once you have these two combined then you have something extremely powerful um which is in some sense one super unified chain. We have uh the L1 that we can scale dramatically uh and still have decentralization again thanks to snarks and data valid sampling. But we can also have things like native rollups kind of be extensions of the L1. And these you can think of them as being programmable synchronous shards. So synchronous because they inherit the network effects and programmable because they can be

fully customizable. And they're extremely flexible. So you can have as many as you want. You're not limited to some number like eight. Um you can customize them and change the sequencer, change the tokconomics, change the governance. Um, and they will basically inherit the same security as L1. So, so far not a lot of fights. I would say a lot of things working in common. Uh, what's your point? All right. So, roll-ups suck. Okay. Okay. That's that's the spirit. That's the spirit. Maybe who who here likes to bridge between roll-ups? Who who enjoys finding which rollup their smart contract is on and if their network is connected properly on the MetaMask or whatever? person. Okay.

You're enjoying it. This is this is you are like this is Yeah. You feel like you're doing work. That's Merrick from LTB by the way. Yeah. So, so the idea of sharding is that the the scale should be an implementation detail, not the userfacing concern, right? It's, you know, like you should be driving on a highway and you can, you know, switch lanes and not think about kind of the the road versus, you know, right now Ethereum feels like a lot of, you know, side roads and kind of exits and entrances and bridges that you can try to find where you're going. Um, and so I mean, yes, the final design I think we agree, although I'll I'll challenge you on a few things, but the question is what is it now, right? like the

current experience right and the you know past few years of building out the current ecosystem right is extremely sub-optimal uh it's really hard to use the fact that like your address space is effectively fra like fractionalized it's actually doesn't lead to scaling it leads to you know a deployed across every every rollup right instead of having more capacity for a you have more a split liquid liquidity. Now on top of this we have kind of like right now we don't have any you know phase two rollups. So we don't actually have any security and the problem is this centralized roll-ups opened up

effectively overtone window to just say decentralization doesn't matter like I think actually this is the most major failure of Ethereum is that it kind of allowed people to say actually you know hyperlquid yeah it's good enough right multisig good enough right like that that's actually the biggest failure I think we've seen over past uh kind of few And so now we don't really like now when you're building a product if you're trying to be decentralized people like why right so that is I think the biggest failure so we can then zoom in into like where the kind of uh beam chain is going but yeah

I mean I agree with you um that the the status quo is imperfect um having said that you could think of it a little bit like democracy right it it it kind of sucks but it's the best thing we have. Um, and the good news is that as you said, we we kind of agree on the endame and the endgame is amazing and the status quo is just the stepping stone to to towards that end game. Um, you mentioned security for example, it is true that the security is very suboptimal today. The main thing that is really bad in my opinion is the security councils. And why do we need security councils? because there is the possibility of bugs either in fraud proof games or in snark circuits and this is exactly what the native rollup

idea is trying to fix to remove the bugs remove the security councils and even more than that stay in sync with the L1 EVM so that you don't necessarily need governance to maintain your L1 equivalent uh rollup. Yeah. But but so the question there is why do you need native rollups? Why don't just go sharding deliver the same experience you want from native roll-ups. Yeah. Uh and I'll I'll challenge you on the you know everything is synchronized part. Yeah. I mean what we've seen empirically is that u people like rollups. Um so we have again please hand who likes you know exiting and entering

rollups. We have one person who likes a single feeling of a single chain. you know you don't need to switch you know all the applications in one place your wallet has all the assets who likes that more so empirically we see more people liking that okay so has doesn't exist yet but okay it's a quick question it does but yeah like has Coinbase Kraken Robin Hood Deutschbank Sony like have they launched have they deployed on shards or are they deploying rollups so again I as I pointed out We as a stadium opened up this overtone window that it's fine to just run a singular server right you effectively saying cool now this big

companies which again it's good for on boarding more people but at the end they're just running a server like same as they've been running like Coinbase is a server in the first place is a server run by Coinbase so it's I mean effectively expectations are the same right we're not we haven't really like solved the decentralization and like the self sovereignty that the whole point of you just need to be patient. Um you know Coinbase has been hardening their their chain continuously stage zero stage one um and I think uh L2B at some point will add a stage three or some sort of native stage um once we truly do inherit the security of the L1. Um, so yes, there is

this kind of awkward adolescent puberty phase that we're in. Uh, but I think we're going out of it and the the future is extremely bright and even though it's imperfect, it is the most pragmatic thing to to be doing. uh you know in the pre previous debate there was this um kind of claim and I think you were telling me that you wanted to make the same claim that uh you know the whole strategy all along was to just let the L2s do all of the work and then kind of rug them by reusing the technology. Well, it it you're trying to say yes. The end. Well, what I'm trying to say is that we had no choice, right? So, the the L the L1 cannot use technology which is not

mature. And the technology just took 5 to 10 years to mature and now that the technology is mature, we can use it and we can have the L1 be superpowered. But one of the things that we underestimated in 2018 is the the power of rollups from a BD standpoint, right? Like Ethereum doesn't have one BD entity, which is the Ethereum Foundation doing BD. It has dozens of BD entities that are going out there and getting deals like Robin Hood onto Ethereum. Uh, and that is an extremely powerful concept. And what if we can have both the security of L1 with native rollups and the BD aspect that is a superpower we didn't even know existed.

All right. So then let's talk about native roll-ups, right? I mean your claim effectively is that hey we didn't like the original sharding or you know some versions of sharding because they're effectively a synchronous composibility between applications and native rollups you claim are going to bring synchronous composibility between well that's base rollups. The native roll-ups is about security about removing the security councils removing the the possibility for bugs in snark circuits or fraud proof games. All right. So this is just like everybody runs effectively the same code like EVM like ZK circuit is baked in. Yeah. But the important thing is that you can program it. So it's programmable role level. You can change the sequencer

change the tokconomics the governance you want to do public goods funding like optimism whatever you want to do you you you can do it. All right. So which I mean you can do all the same launching on a smart contract chain right you could just design your token economics and I mean sequencing is less right but sequencing is where you know I thought we all brought censorship resistance and so if you're outsourcing your sequencing effectively giving up on censorship resistance you can't always just launch as a contract because you know you might crowd out other applications and there might be like gas fees dynamics and and things like that sharding solves that because you can always allocate contract in its own shard. Well, not really. So, right, you have I believe um an EVM

contract on one of your shards and that one contract is you know big and vibrant and it cannot itself be subsharded. So, EVM no but we are the EVM. The EVM is the winning thing as a de facto standard of the industry. Well, Salana would argue a different point here, but I'm not from Salana. Look at that. But, uh, I think like if we're talking about, you know, 100 million user applications, right, which we are trying to bring as a ecosystem like EVM will, you know, doesn't scale and not going to scale in next few years toward that and so you will need did you watch my talk yesterday? H so yesterday I gave a talk gigas L1 we are going to scale

aggressively specifically the plan is 3x per year for the next six years. So that's 600x over six years from 14 transactions. Um no we're higher than 14 transactions. Um we're closer I don't know to 40 or something. Um but it's it's going up. Um and the hope is that it's going up faster than um demand for for block space. All right. And so let's go back to what you said about synchronous. Uh yeah. So that's with the base rollups.

When you share a sequencer and when you have real time settlement through real-time proving for example, you unlock the ability for contracts across different shards or or rollups to have flash loans for example. So but but when you do this you're effectively limited by what that sequencer can process. So you you are kind of bringing it all back to a single machine and its limits. Yeah. So one one of the things that you know we've seen empirically is that the block building is inherently a sophisticated thing. Um and um there is a lot of centralization. Um the important thing that we're trying to do is make sure that even though the block builders are centralized, they

can't do anything bad. So in particular, we don't want them to be able to censor. So we have things like fossil. We don't want them to be able to cause lieness faults or safety faults or or any bad thing. Um and so we're basically leveraging the good things of you know powerful builders which is that they can find value across multiple disperate places of the ecosystem and create these synchronous transactions um without the downsides of of of censorship and security. Yeah. So sorry to interrupt. I mean you are making my MC job super easy. So, thank you for that. Uh, I hope the audience is uncertain. It was just a quick reminder that we might have something like two, three minutes left.

Uh, so don't forget to scan the QR code and vote if you didn't already or if you arrived late and I will let you keep fighting. Yeah. I mean, I think like the the challenge that I have with this approach, right, is you're effectively saying, "Cool, we're going to have like how many right now? Three sophisticated blog producers." U uh I mean if if NEO gains any like significant traction then um it will have the same kind of centralizing forces. Actually we have 30 seconds. Sorry I just missed the big timer right there.

Please go on. So what will happen a a natur like a natural evolution for mature chain is to adopt what we call PBS proposer builder separation. We have something called meth boost where basically as a proposer you don't actually do the work of block building you outsource it to sophisticated builders and this is a perfectly permissionless and natural evolution and well not if you're doing this at you know few hundred millisecond or even lower time frame there's just no time for somebody else to try to propose and communicate blocks to you and this is already happening even on Ethereum where uh if you want to get inclusion into next block you have relationships with all the validators and you effectively pushing from your darkpool transactions directly uh into this. So the reality is

like yes there's now like a whole ecosystem of projects that are kind of working around all of the Ethereum you know sophisticated root goldberg machine uh to make sure they actually get the performance that they need to build applications and absolutely sorry but we are done so I think it was a super interesting talk mostly about well uh if we set the technical uh aspect aside very interesting call about very interesting fight about when is the right time to do thing and what is the speed of R&D and how do we manage to compensate basically well censorship resistance another thing so thanks you for this guy I think you can continue you can keep fighting after because there are still a lot of going and yes yes yes then take it in real life

because you're here okay and well it looks like you won this stage but I feel like it's not the it's not the last one so congrats congrats to you both thank you Nicola thank you Thank you.