Panel I Rethinking Data Availability | Pragma Istanbul 2023

[Music] we got four more panels to go we're going to start off with our fourth panel for the day for this panel we we're going to talk about data availability uh there's a lot of stuff happening in this space a lot of optimizations a lot of changes and maybe some disagreements too we'll find out very soon but to talk about data availability I want to bring on three amazing people I want to bring on Shon from Igan Lair Ilia from near and Nick from Celestia we're going to talk about how and why this is important and the direction we can see things going and some things that we can look forward to in the future so without further Ado please join me in welcoming Nick sham and Ilia good afternoon all right I'm gonna try to set up the the stage here by doing a brief intro on why this is important and what it is so we'll kind of go in in order first just so we can get different perspectives but U I'd like for the rest of the conversation to be more of a actual conversation between all of you versus uh one by one um so let's start off with me asking you Stam uh why is data availability important data availability basically is the following primitive so in the in the ethereum modul road map we offload the computation out to rollups which basically do the computation and then create proofs that the computation is done correctly when you do this it's important to be for anybody who's observing the system to be able to access either the inputs or the outputs of the said computation because once you have it you can verify that the computations were done correctly or continue on when the sequencer is no longer around so this is what gives you the unconditionally correct trust model of rollups so we need a place where people in this case rollups are able to publish data which is universally accessible so that's why I think it's very important and it's important for this this primitive to be highly scalable because all the rollups uh the the total throughput of all the rollups is constrained by the total amount of data that can be returned so this acts as a fundamental constraint to the entire scaling road map sir what's the Nuance here so it's constrain on the total amount of data being stored on the layer one or Al together because it's a bandwidth problem or something yeah so if you want you know the the the total data that is return returned by all the rollups is constrained by ethereum's data throughput if you wanted the exact security guarantee of ethereum so this imposes this constraint like you know that this the toal total amount of scale is now constrained and that's why it's expensive to use rollups today to an extent because they're all paying the cost of data so by scaling data availability and reducing the costs even at a slightly lower trust model is still going to be highly beneficial for a lot so maybe IIA if you want to add on to this how does it actually help with scalability yeah I mean sure I'm kind of uh outline the basic idea I think maybe looking a little bit from a different direction I think users and you know to many extent developers don't care about specific blockchains right they care about actually being able to we're we're going to start off with the spicy stuff I'm all for spiciness here so yeah that's why you invited me uh and so the they care about the value they get from it right you know be that access to kind of like liquidity markets be that you know um kind of assets and participation with other be that you know gags and so the kind of from this perspective the when we talk about scale right like what we talk about is you know making this more accessible to people Mak it you know fees lower uh and at the same time being able to what I call unify security right what you want is that as a user you should not need to think about like oh I'm going to BNB now this set of potential failures that can happen to my assets right like that should not be something that this user and to some extent developers need to care about and so what rollups kind of introduced is a primitive right that uh enables that where you don't need to think about the failure modes of um of the kind of rollup operator which you know is another chain potentially and you can rely that there's a way to recover from the fa the kind of the failures of the operator but in turn you know if indeed if you want to get same kind of level of security or the maximum level of security we have right now in we3 ecosystem which is ethereum you do need to pay a haly fee right which is way higher actually than many of the scalable blockchains in the E system and so uh so the question is how do you get best of the Both Worlds how do you get way lower fees uh and at the same time have you know as much security as possible so that as a user you don't need to about this details at all and you can just use applications kind of in unified View and so that's where you know scalable data availability that has a fair amount of security is extremely important and linking it back to ethereum as well is an important primitive as too and and maybe kind of just to to round it up Nick how do you kind of Define modularity in in this kind of theme which pieces here matter for data ability versus kind of being able to actually create more module architectures for different blockchains to do more than just High thots so I also just want to touch on like the importance of data availability is because it's secures the whole ecosystem built on top of it of layer twos and rollups um and the issue is that um data availability in the current model in which blockchains are built is just not fundamentally scalable um and so the to to to my knowledge the only actual solution that can both uh scale that can scale data availability in a secure way is data availability sampling so hopefully we talk about that a little bit more um but how does data availity play into modularity is that uh in a modular blockchain you one of the insights is that you can decouple data availability in execution so you don't have to execute all the the data that's being processed by the L1 and that enables you to just uh scale block size uh and not have state blow and all these things that have been sort of bottlenecks for scaling blockchains and instead push that off to the the upper layers and this gives developers more flexibility as well as they can Define their own execution environment because they're not limited to the evm or the slot of virtual machine or whatever virtual machine the layer one blockchain has so modularity and and data availability go go really hand in hand in this new model for how we should build uh blockchain applications since uh already kicked us off with the with that comment um I agree users don't care about things like this but developers definitely do so what do they need to know about this new world and and kind of what changes for them so developers care about few things one is they want to launch to and have access to as much users and as much liquidity or whatever kind of the related Market distribution yeah distribution yeah broadly speaking and so because of this they want to have you know the security guarantees that if they launch that their application will be there it will not go away it will not disappear because of somebody else um as well as that you know as users you know coming in they have this guarantees uh around security and then on the other side uh this goes now into two into a fork right which is you want to launch your own rollup and then you have a set of you know criteria looking at or you're launching on a you know some other chain or roll up and you figured out like what what the parameters are um and there's like a very you know interesting decision making right now happening which uh you know for a spicy take usually depends on the token launch uh structure um and so I think like at the end what you're looking for is you know the cost overall right because I again ideally you hide the costs from the transaction costs from the user right right now you cannot do that because they're so high but you know if they're for example you know on near right now it's fraction of a cent it's like one10 of a cent this is just a transaction cost normal transaction normal normal transaction cost on near Isa like tens of a cent right so you know a lot of applications just hide the transaction cost completely and so for them it's just an infrastructure cost same as like AWS or Google Cloud they just like pay you know like stake near use the proceeds from staking to pay for transaction fees and then they you know generate revenue on this and you know it goes into toen or whatever the model is and so yeah so that's kind of the one really important piece and so that it defines you know the diability cost the settlement cost like all of those pieces then you know get aggregated into your final transaction cost uh and so for that you know we can discuss like what are the uh cost structure for daa Ability what is the um you know settlement speed all those things that like technical parameters let's let's get into it um maybe maybe I think the the last thing that'll be helpful for everybody else here to know would be just understanding what your D solution does so let's let's talk about what IG layer is or how you kind of think about yourself in the landscape and then I want to do the same for near and Celestia um yeah so we are building a data availability solution called igen da which is built on our reaking platform called igen layer IG allows ethereum stakers and node operators to stake and participate in a variety of different Services any in you know distributed system Nick here was alluding to modularity in our vision the number of modules that we need and usually when people talk about modularity we think about like you know two or three modules like data availability and consensus and execution in our vision there are may be hundreds of modules that people may want to build you know AK to the software as a service era in the CL cloud like people build many many hypers specialized layers which then concatenate to offer end users and applications the set of properties that they want and when you have many many modules what is very important is that if your end application uses tens of these modules you know like an end user application in Cloud today uses 15 SAS services on an average if you use 15 of these modules then you are as trustworthy as the least trustworthy of these modules because the attack is always like that what is the weakest link in the chain so our solution to this fundamental problem is shad security the idea that there is a common security a common trust layer coming from staking and node operations which can then Supply trust flexibly to anybody who wants to innovate on these modules so one of the modules that we are building is to solve the data availability problem called Igan da I da is built from first principles as an adjacent data availability layer to ethereum it's not a chain pretty much everything else you see out there is a chain and a chain has its own like constraints and structure whereas what we are doing is simply what is the minimal adjacent to actually scale the ethereum data availability road map and what we do is take some of the best ideas looking ahead in the ethereum road map like d Shing proof of custody the crypto economics around it bring it all into a protocol called agend which we can ship within the next few months so that's that's our our view and because of this particular like angle which is that we are not building a chain when you build a chain you have to build a consensus protocol you have to build lots of other things you have to build a peer-to-peer Network you have to gossip there's lots of complexities which actually drown out any scale that you can actually achieve our vision is very aligned with the ethereum vision which is nodes have very low capabilities but together they all have a lot of capability so it IG da is built on the principle of exact horizontal scaling if each node has you know one or two megabytes per second and I have hundreds of nodes the net throughput available is hundreds of megabytes per second how can we build a system that actually leverages this and achieve the maximum throughput possible and there's there's lots of principles one of the principles for example is no node should ever touch all the data if you want a system to to be exactly horizontally scaling no node you know either inside or outside the system needs to touch all the data so we're building a data availability with a road map to scale ethereum data through put you know to hundreds of megabytes per second so that's that's what uh I is all right so I mean definitely agree on no no should touch all the day data and so near from the start um have been building a sharded blockchain to pretty much scale both you know execution but as part of the execution you need to scale data ability and so the nearest consensus itself actually has data availability weaved in into it so when the first step when transactions are sent they actually are AAS a coded that sent out to other validators uh in kind of partial way to allow kind of everyone to pretty much start a testing already on data availability for the next block while the execution kicks in and so this this is always was near design and so when we saw kind of there's interest in this way of leveraging uh you know running rollups with data avability it kind of was pretty clear for us that we can just offer what we already have um in the in the near ecosystem and so near right now I mean uh is about 4 megabytes per second uh per Shard and so we have four shards so 16 megabytes per second of data availability uh there's a plan to expand number of shards with some of the launches in q1 um so that kind of allows to have a you know pretty sizable uh so what's the tradeoff that you make when you expand the number of shards is it just is it finality or or kind of data availability altogether so so right now is the limitation has been that um we have so we haven't enabled uh so-called State transition challenges so this is the same problem that everybody else in the space has nobody wants to do challenges on state transition optimism poka nobody wants to and so uh kind of a few months ago we actually switched to implement stateless validation so similar actually looking forward to sell the ethereum road map and so we're planning to launch status validation with that we can actually remove the current limitation on we have still some validators at validate all shards right now so we can remove that um going forward from q1 and so we can start expanding our shards now there's you know still limits on how fast you can synchronize until you start needing to do hierarchy uh but I think that limit is probably like 100 shards um is uh at this point and Nick I'll let you talk about celesa as well and then we'll move on to other questions yeah well what um sets Celestia apart from igda and and near is that uh we've built uh and implemented and launched what I think the research Community agrees is the only actual way to securely scale data availability which is data availability sampling the reason for that is that there's there's a few things uh at the end of the day blockchains are meant to be verifiable systems and uh in order to verify da you either have to download all the data yourself to hold it in your hand and know hey the data is actually here someone published it it's public information or you can rely on a committee which is sort of what uh I think to my knowledge what end da is doing uh and and kind of uh indirectly what I think near would be doing or you can have a system where you uh eras your code the data and then serve it to clients who can sample randomly from it and they can have a probabilistic guarantee that the entire data is available um while doing while sampling maybe 1% of the overall data so it reduces the overhead of verifying da and that to our knowledge I think that's the only uh Solution that's so far been been proposed and so I think um yeah I guess what set Celestia part is that uh it's actually like really trust minimized uh data availability from from from the start and you're not relying you're not trusting a committee saying hey trust me the data is here that end users can directly verify and and the be beautiful thing about that is it enables this new kind of related to what Trum said about not everyone needing to touch the data it enables this new property that no other chain has had before which is it scales with the number of users the number of nodes in the network um and so there's no fixed block size for Celestia it grows as more people join and start to verify and Sample the underlying block um some disagreement here uh data availity sampling actually you know Celestia Founders are one of the pioneers of data availabity sampling and is actually a really cool technology but it only helps when you are a rollup on Celestia so you have to build a native rollup on a chain and then if all the validators of Celestia collude and make the data unavailable you as a user of that rollup can sample the data and know oh yeah the data is not available that's what you know Nick called self verifiability I think this is a really cool property I want to emphasize why we didn't actually go take this path because if you're building something on ethereum everything is just a committee cestia is just a committee as far as ethereum is concerned ethereum doesn't cannot do data availability sampling on Celestia ethereum trust that a majority of Celestia nodes said that the data is available and then makes progress on the rollup contracts so because we built our protocol to be native to ethereum and as an adjacent to the ethereum rollup road map we decided to make data availability sampling a secondary feature why do I say it's a secondary feature there are some features we have to be which have to be built in protocol it has to be inside the protocol to actually for a feature to for example nobody can go and change the ethereum block size but you can build a new wallet by yourself it's permissionless data availability sampling on I is exactly the same thing anybody can build a sampling mechanism on top of the igda protocol because it already has Eraser codes I think you can also build it on the near protocol because near also has an ARA code ARA coded commitments is what enables data availability sampling but once you have it in the protocol anybody can build a sampler out of protocol but from ethereum's Viewpoint everything is just a committee there's actually um an approach to have data ability challenges on ethereum ah yeah you know there has been some recent talk un announcement about it so you know I think this is one thing we may all agree um the idea of a data do you want to talk about it Nick seems well some someone explain it and then I'm just going to tear it apart okay let me explain it and then it can tear it apart okay so the idea of a data availability challenge is somebody makes a claim that the data is available and if you don't agree just like in an optimistic rollup somebody makes a claim that the execution is X and if you don't agree you go post a bond and then you run the execution on chain and check whether it's correct or not data availability challenges extends the same idea to data somebody claims that the data is available you disagree you go and raise a challenge on ethereum and see whether the data is available or not that's a data availability challenge I'll let Nick tar it apart uh yeah well the the problem with data availability challenges is uh something that uh the research Community has known since 2018 uh it's called the fisherman's dilemma essentially uh it relates to the fact that data withholding is unattributable so for an outside Observer when someone withholds data then you can't actually objectively prove that data is withheld because that person can always just later reveal the data so now you're you're stuck in this situation where anyone who try who raises the challenge can uh basically be triv trivially uh disproven and it relates also to another thing which is that um and which is kind of related to what Shone was talking about of everything every da solution external to ethereum is just a committee that's true and what I would add though to that is that any uh committee that is or any da solution needs to have a native asset to give it any cryptoeconomic security because same reason why data challenges don't work uh you can't prove on chain that someone did data withholding so you can't slash someone for for example if they stake eth on this da solution you cannot prove that hey you should slash this person's stake eth the only way to slash for data withholding is socially and socially slashing slash uh collateral has to be native to the data protocol and which is why for example Celestia did not pursue some sort of like uh E3 staking type of way to secure itself it actually merits and requires a new token uh to to to do that so yeah but I want to add that that has nothing to do with data availability sampl near protocol for example you know if data with was withheld on the near protocol you know a majority of validators collude and uh they make a claim to the ethereum protocol that the data was available the near protocol will stall you know the blockchain will stall and then the validators will coordinate a social Fork forking out the near validators who made this wrong claim and therefore they'll be slash so there is a cryptoeconomic guarantee but addressing the cryptoeconomics of I do you want to add anything on this well I think the interesting part there is is our partnership to actually uh plug in near with ien lir such that on top of the near security which is whatever like 7800 Million worth of crypto Economic Security on indeed that like pretty much if at least one one six of validators are honor like there will be actually recoverable data across the all shards on top of this then we can re kind of Leverage reaking on ethereum to have the reaking validators to pretty much sign on like and validate as well the rasure coding and commit that signature as well and so this actually gives us more security and From ethereum perspective actually unifies as well that that you know it's using very similar underlying security which is why the the spicy question is is it near than the [Laughter] I okay so the I I'm going to address both the things together what is the cryptoeconomic security of igda so Igan layer is this reaking protocol and the way it works is you stake eat and then you borrow Economic Security but what does it mean to borrow Economic Security what is this like new object it's basically saying that if there was a violation that happened on your protocol if you can write an onchain slashing contract what is a slashing contract it's specifies the conditions of misbehavior under which your e will be taken away if you can write an onchain slashing contract and then you can basically um slash somebody's eat and what you're borrowing is economic security means that if your service went wrong you will be able to slash this amount of eat so unless um so that gives you you a certain amount of guarantee an economic guarantee that your service won't go wrong okay so now coming to Igan da and why it's not so simple to borrow Economic Security for aan da which is a point that I think Nick is trying to allude to is you know in when you're running a data availability system there are two things you're doing as a data availability system number one you're downloading and storing the data because it's data availability but you're also serving the data these are two things you have to do and what we have with IG da is a mechanism for slashing for data not storing data it's called prove of custody so this prove of custody is a mechanism which basically ensures if you hold the data then you have to do certain kind of computation with the data and then and along with some private secret and then post a proof you know directly or indirectly that you've done that computation so if you're not storing the data you will get slashed but the problem is that if you you may store the data and not serve it right so if you may store the data and then not serve it why may you not serve it there may be two reasons one is you're lazy and you don't care and you don't serve or number two you're just colluding to like adversar withhold data and so here's where the trust assumptions of different protocols have there's a lot of nuance so one has to think about it carefully in is constructed on the Dual pillars of Economic Security and decentral ization it's not a purely Economic Security protocol what do I mean by that when you rely on economic security you're only caring about how much money is sted do you have 100 million or you have 1 billion it doesn't matter whether one guy Stakes 100 you know this 1 billion or thousands of people each take 1 million each because if that thing went wrong I'm going to be able to slash 1 billion so this is called Economic Security there's another pillar of blockchains if some of us have forgotten it's called decentralization right which means there are many there are many distinct operators switch to one server now for everything so the idea is that in decentralization there uh there are many many operators each of whom are you know should be able to operate with very limited resources many operators and they are all you know independent and therefore are unlikely to collude so I is built on the assumption that if that is the case if you actually have decentralization which is lots of node operators who can actually uh don't Cod then you get the following mechanism which is that if you want to serve the data like there is a competition among the different notes to actually serve the data because data is widely dispersed no one guy holds a monopoly on the data because the data is a coded you can download it from any 10% of the notes depending on the rati ratio in nearest case any one6 of the nodes so it's it's a parameter but what that does is as long as everybody 100% of the nodes don't collude together there is always an incentive to serve the data that's the cryptoeconomics of igda because it's nuanced and has all these things we think of a as a special service and whereas Economic Security can be easily borrowed for any other kind of service D availability has to be much more nuanced and so that's that's over so I just want to add I think one of the the third and probably most important pillar for for blockchains is verifiability right like uh sure you can have a committee and make it more decentralized but you're still making an honest majority assumption on that committee and that is not what like why we're building blockchains to get out of that kind of system uh otherwise it's it's similar to web 2 where you can just have a Consortium a bunch of companies like sign everything and it's like hey we trust that that's not what we're here for we're here for verifiability by end users and so that's like that's again the only solution that can give end users verifiability is data availability sampling you get the benefits of data availability sampling you have to be a Celestia rollup you cannot be an ethereum rollup and then use Celestia for data availability and get benefits so you know that's a real power of having an integrated data availability system so I said like I layer I da is designed as an integrated or like an adjacent data availability to ethereum cesa is designed to be very powerful you know this special property of data availability sampling because it's integrated with consensus that's what gives it the superpower that actually if data is unavailable the chain will Fork along with it and but unfortunately this this benefit doesn't translate when you're in AUM roller um so uh my next question is uh what is something that you disagree on I thought that was the whole panel well actually most of the time when we hear these things um one thing I've notied personally is that uh a lot of this talks about the the economic security of uh the protocol which is kind of bundle into a binary will it exist or not will it fail or not uh but this particular thing is espcially important to developers on what exactly is being stored and how does that compromise anything that I'm building uh if I'm building a company or protocol anything else on the successful side so kind of forgetting the the cryptoeconomic piece of it um can you give us a way to think about what does it mean to cost sample and just actually store that information and how do I think about the cost for storing this uh whether it's my rollup data or my kind of app data that's being kind of positioned into Celestia or or I or anything else uh how should I think about what is going to cost me how does that change over time who's paying for this who's providing these storage Services uh what's that in incentive model and uh and kind of how does that evolve I mean I I'll take a spicy take here um I think the reality is we have to explain what it is before you kind to make any comments like I want to understand what's a good intuition for how much does it cost with a solution versus putting call data on ethereum uh just just so we have some in a way dollar figures or even like multiples um I'm not sure exactly what the per megabyte clost is like a putting posting data to ethereum but uh earlier today on Twitter uh there's a team building a game on Celestia and they did a comparison of like if they do it on ZK sync it was like 20 cents a transaction can't remember they went down and then Celestia they even actually running on salana was like on the neon evm was like .5 and I think Celestia was 08 or something like that so uh like orders of magnitude uh less expensive expensive as of now and hopefully actually cheaper because we're planning to decrease the the the the sort of Min fee that we've set so I think one of the beautiful things about this is that I think the could you define what the fee is so the fee is for what particularly onetime cost is it a rent right so okay first of all it's really important that people understand that storage is the wrong mental model to think to use when you're thinking about data availability it's more about publishing and P something that you publish is just something you pay once it's like I want to it's like adding putting an ad in the classifies of your newspaper like it it goes in it's there and then it goes away uh it's not meant to stick around so there's not like this recurring cost of like oh I need to keep the data around indefinitely in fact like all data availability Solutions have pruning like built into them for that reason um and so uh I can't remember what oh yeah I think one of the cool things about this is I think that cost has been a major blocker to uh scaling obviously ethereum l2s and there's a lot of use cases that have not been built and and there's this idea in economics of induced demand where you know imagine like electricity what costed like I don't know $50 per kilowatt hour or something all these use cases like running a microwave or I don't know charging your electric vehicle or whatever would become uh infeasible or just not not worth it and when you drop the cost of electricity to let's say like you know a cent per kilowatt hour all of a sudden people are going to use a lot more electricity because all these new use cases get uh become feasible and useful and so I think that's what I'm excited about for Celestia launching blob stream which allows people to run ethereum l2s using Celestia SDA all the other other da Solutions will uh unlock all this like pent up demand and and use cases that have been held back uh so I think and and the end user saves money I also think the developer can reap more of the execution fees for themselves so the developers can actually like sustain their business yeah so to second that we're 8,000 times cheaper than aium so that's a big number yeah uh so where you're right now your base or op or arbit room is like 4050 cents we are yeah 4 zeros one something like that uh or 5 Z one and so uh the kind of I think the interesting thing here what happened that that's what I mentioned I think developers will just cover this fees and like on near again you can just cover it from the staking reward the real costs now kick in and the real costs are indexers and all of the infrastructure you run around this to actually serve this data to the users to properly interact with your apps and I think that's where we kind of forget like yes data availability is the publishing part it's making sure that you know people can verify can run the indexers can run all those infastructure but now you know if you want to spin up Easter scan for yourself it's 500k a year Min right and that's a real costs so so I think the important part here is to think not just about the blockchain costs now especially now uh but actually about the full cost of developing apps here and this is where sometimes it makes sense for you to launch a rollup and and you will have like two three four million a year of like extra costs on the services and sometimes you should use a layer one that has all this infrastructure it integrated and uh kind of you don't need to pay this extra cost so that that's where I meant like there's a decision actually Fork there around like what the ecosystem You're Building how you actually structuring your application and what are the other pieces that you need around your infra anything to add yeah um one of the one of the things we think about a lot is the economics around data availability um the the um the principle that we are building out of is actually uh you know people think a lot about block space as a kind of fundamental constraint we're building for a world where like block space is not a constraint and to to have an Avoca of analogy you can think of like is there a concept of cloud space like if you wanted more like compute AWS is will create more compute for you right so there is no concept that this is the cloud space limit there's no Cloud space limit Cloud will stretch itself to accommodate whatever is needed and one of the mechanisms one of the important me economic mechanisms by which this happens is reservation right if you go to AWS you can either have a spot instance can randomly go and try to get an instance at the moment sometimes there won't be instances but you can reserve an instance it's just for you for the next one year I da is built on this dual economic model of spot bandwidth and reserved bandwidth because as a rollup you know you want to have the cost basis of your rollup to be more predictable right you're running an aeroplane company you want to know what the input costs are so that you can charge your users so it's the same thing when you're running a rollup company so I is built on this as one of the principles which is that the you can reserve data and this allows us to provide more data because more people are using it so that's number one number two we have a mechanism where people rollups can pay for reserved data bandwidth in their own native tokens without taking exchange rate risk right if you're in layer one one of the really cool economics that you have is you pay a certain fraction of your token for like all your validation but if you're a rollup you suddenly like have this like obscure economics where you know know you cannot say that you're only paying a fixed fraction of your own token for data availability so igda is built on res reservations through native tokens number three I is built on Dual staking so you not only can stake e but also provide additional utility for your own rollup token by allowing that as a stake staking token aund allows this as a native feature all these features are fundamentally economic features is on the data so maybe the last question here which uh hopefully we can make some quick predictions on U how do you kind of think this this economic model pans out I mean is is just going to be subsidized for other potential Revenue sources everything becomes free because everybody has to then take on the the indexing cost which is the real kicker or or are we saying we kind of find a stable equilibrium here like what is like two years from now data availability look like I mean I think the reality is the cost will go down to to whatever the compute storage costs are yeah a little bit it gets cheaper every two years yeah and so I think that's why so for example near why I'm saying you can just stake and you know pay from staking rewards is because that actually aligns the cost structure and we have actually a way specifically where you have your own token you can just distribute this token to the stakers who are staking on your node that pays for your transactions so you can be literally same same idea you can be in your token economy and pay for transaction fees under their knes and so Aurora been working like that as well where they they cover transaction fees for first for like first 50 transaction fees for users but like that that's model like kind of just built on top of as you said not not embedded in prodal but just build on top in general I think we'll be going forward where uh we'll kind of start moving away from direct transaction fees by by the user when you have like again we're talking about you'll pay something like you know $1,000 per like 10 million transactions or something like that and so like you you'll need a you'll need a lot of transactions yourself right um and so we already have this on near even like uh with uh you know we have an app that has like 700,000 daily active users right and they cover all their transaction fees from staking so like that's totally total work on a Model um so I think like yeah the M the the main costs will start moving up to like actual infrastructure that is that does need like a lot more capacity and you know with blockchains I think we'll see kind of compression fees pretty dramatic any other quick comments on this I there's a a blog post written by Chris brisy of placeholder uh from back in 2019 where he just talks about like the economics of of protocols generally and um he has a quote that I really like which is that protocols should strive to be uh minimally extractive but globally produced and consumed and so I think uh the goal of a data availability solution is to be uh to produce at scale and sort of like sell block space at massive scale and you know hopefully almost at Cost be like minimally extractive or like not uh sort of like unnecessarily extractive I should say and I think that's kind of the uh the the philosophical alignment of Celestia is like it's more like uh kind of like AWS it's like AWS doesn't win by having only you know 10 computers that like or like a tiny server data center that everyone has to compete over and price gouge people it's like no let's expand capacity as there's more demand and just make sure that there's uh know some amount of margin that is is charged to make the whole thing economically sustain awesome that's a perfect to end on um nickia Shon thank you so much for an amazing conversation thank you thank [Applause] you here you all right up