Splitting the Block - Daniele Palombi

hello everyone I'm Danelle from 20 squares and uh so for those who don't know 20 squares we do Consulting on uh like economics matters in a team uh from mechanism design Market design auditing of uh economic mechanisms and all uh kinds of uh interesting R&D uh on economics Mech mechanisms so today I'm going to tell you about something in the context broad context of prec confirmations as anyone here ever heard about prec confirmations okay that that's not important uh we're we're not going to be needing that but yeah so okay splitting the block what do I mean well first let's have a little introduction and plan of the talk so the problem we're trying to address is that currently block space is allocated monolithic Al so basically uh after the merge under PBS uh a validator can auction off the possibility of uh building their block uh to other external third third party Builders and uh so the the whole block is sold over to these Builders uh we're going to explore if this is the actual best way to allocate uh a a block or a block space and uh we're going to explore our proposal for an alternative design actually this is a bit more than a proposal uh so this is uh work we have done with manifold finance and uh basically XGA the product that's spun out of it it's uh it's been on testet for a long time and um we also have run some experiments on mainnet be back in like very little time okay so and we'll also see if I mean well go through the mechanics of this proposal but also uh see how this can be implemented and we'll see that this can be implemented without a lot of uh changes on the like protocol level so that that's good because this means quick adoption uh okay so let's first have some preliminary takeaways about the current state of block building in ethereum so there are very few dominating block Builders like uh we're going to see in the next slides that like two three um dominating block Builders actually build like 90 plus per of the blocks that are built on ethereum uh which is concerning on the centralization side but I mean uh PBS proposer Builder separation was actually I mean the V is that in order to prevent centralization also on the validator level so the initial idea by flash BS was okay Builder centralization is going to happen anyways so let's separate Builders from validators from proposers and so the validator set is going to be uh I mean decentralized as ever and uh yeah and like Builders are going to centralize and whatever but but maybe this wasn't a good idea uh in Insight um also uh the infrastructure has become very uh concentrated I'm not saying centralized is more concentrated on the type of tools I mean also centralized a bit so uh for example uh most of the validators nowadays run me boost and uh so it's like 90 plus percent uh and uh a core piece of the MB boost infrastructure is relays and there are like six dominant relays and like the top two actually built more than 50% I mean relay uh more than 50% of the like actual blocks that end up being built and also relay don't make a lot of money uh actually most relay make relays make none uh and I mean having a core part of the infrastructure uh which is expensive to run basically make zero money it's I mean doesn't seem so stable in the long term and viable uh we have also some problem problems on the user side uh so one is uh at this point the interface between L1 and L2 is a bit confusing for users that it should be hidden from them uh but we're not really going to address that in full and uh also like user needs are pretty much heterogeneous like not all transactions are made equal and and they don't have equal needs but we're basically throwing everything in the same market so I mean we we also need to address this problem okay so how is monolithic okay how is block space allocated monolithically today so at the moment as I was saying a couple of Builders are uh basically build most of the blocks and uh they usually care about only the the filet of the block the the top part the the the most important and profitable part of the block because either uh uh like big Builders are basically integrated Builder Searchers so they have to run their uh their own uh like hftd routines and whatever do arbitrages uh uh CX Dex uh whatever or their uh like auctioning off on like uh their I mean the the the top spots of the blocks uh to other Searchers competitive Searchers um and nowadays the protocol and the I mean rather than the protocol like the the Meb boost protocol only allows to sell a whole block uh but at the same time transactions compete for Block space okay and and also there's a a huge risk of uh verticalization of actors along the value chain so uh I mean if I'm a very good and profitable Builder and we're already seeing that it makes sense to also uh spin up some validators and then when I get to propose I don't have to participate in the Meb boost auction Rat Race I just can build my block uh only take care about being as profitable as possible and I'm doing actually better than having to bribe a validator to um to to to actually pick my block okay uh also as I was saying not all transactions are made equal so yeah uh for example there are differences in valuations of position and time sensitivity for various kinds of transactions by time Sensitivity I mean like I need to get included in this block or by this block uh by position is I need to get included in this part of the block uh so for example position dependent and time dependent uh stuff is like arbitrages and stuff that usually wants to go at top of the block so for example uh stuff that uh relates to um CX Dex uh arbitrages or or in general high frequency trading but there are a lot of like timing sensitive things like swaps for example I need this thing to happen but I I mean uh possibly in a timely manner okay but I don't need this to happen in this precise block I just want to swap some tokens from for something else or like totally position independent but time sensitive is rollups I I don't care about uh where my blob Bing transaction is going to be included in the block like positionally but I need that to get that in by a certain time because I need to roll up data so users can be I mean safe um so yeah uh if blog space is if not all blog space is created equal can we sell it as different goods and uh yeah so there are various approaches uh that are being explored by I mean these can be all bunched in the like pre-confirmation space okay um so one approach is for example State lock auctions uh which is basically instead of selling like positional inclusion I'm selling um like Priority Access to a given part of state so I'm selling a lock to that uh portion of State uh this is pretty hard to pull off into to price actually so we're not going to go on that but but our proposal is is also compatible with that um what we're talking about is actually selling different like trenches of the block as uh uh different uh like to to different validators not validators Builders and uh I mean by segmenting the blocks in this way uh I mean it could be more profitable like on average it could offer uh like some smoothing for example uh by virtue of like if you split the market then uh and you split the the the builders then for example the top Builder is incentivized on to keep the block as lean as possible uh in order to optimize latency and maybe they don't want to wait too much for transactions to come in uh because they have to to optimize for their their own hft thing right they have to beat for the Boost auction but then for example we can start filling the bottom part part of the Block in advance and so we can put more stuff in we can have different valuations for it fill blocks more efficiently okay um yeah okay this is like all nice talk but uh is this feasible I mean we need to basically set up a different Marketplace for stuff that that goes on not at the top of the block or like that that's different from Meb boost auctions uh also what does this entail for like execution layer and then the the protocol the consensus uh so XGA our design uh actually uh helps with this and does not require any changes at the protocol level uh we have been working on this for quite some time at this point um and let me repeat this is already on testnet it has been on testnet for a long time and we have run experiments on Mains so we're we're basically kind of ready to launch it I mean our partner is ready um okay so but before telling you about like XGA let's see how like blog splitting is happening already kind of today so as I was saying now we have blob bearing transactions so we are separating concerns if I am a rollup and I need to I mean include like roll up my data and now I have a special kind of transaction that I can include uh and they are basically priced differently than like the the usual stuff that that would go I mean normally on the M pole uh so but the thing is since the block space is still sold monolithically I still need to go and talk with the w w with a a builder to get my blob bearing transaction included and this introduces some friction takes away some money from the validator it's suboptimal so yeah now to XJ our proposal is to split the block into two parts top and bottom uh I mean it's it's not important but uh we're fixing the bottom part currently at 5 million gas um and both parts get allocated differently we we we'll see that later uh so I mean like the the top part gets allocated through a traditional MB boost auction and the bottom part uh is auctioned off on a like uh a dedicated L2 that it's it's made made like purposely made in order to handle these kind of auctions and also secondary markets for for this kind of stuff so uh let's see the the bottom part of the market okay so the bottom Market can be implemented in in several different ways but at this point the design is uh centered around selling like the bottom part of the Block in advance so it's going to be so in the form of gas tokens each token represent one 100K gas and this is a a a like a token Dimension we have figured out by running simulations and looking at the historical data and uh this auction that runs on this XGA L2 uh actually uh assigns gas tokens to possibly more than one winner so it's a multi-unit auction with multiple winners it's a like kind of traditional capacity auction for for mechanism design neres and uh so when the deadline of uh the like traditional Meb boost auctions uh comes uh like bottom of block bundles have to be already submitted and they you have to burn your gas tokens that you've bought in advance so it's kind of like Futures it's it's not really but I mean you're you're you're selling gas like block space in advance uh and and in a size the uh I mean like quantized amount and and obviously bundles that are submitted cannot exceed the gas token value that that has been burned uh so uh okay then what happens uh so right now we have to have a dedicated relay for running XGA we're working on a relay agnostic version that's that's going to come in the future but right now uh uh the usual Meb boost winner sends the bundle like the block to the dedicated XJ relay the bottom bundle is sent to the XJ relay by the L2 Ian the the part that that takes uh care of the aggregation of the bundles and then the relay does a bit more work than a traditional relay it it merges bundles so it basically attacks attaches the bundle uh the the bottom bundle that was sold ahead of time at the bottom of the block and takes care of all the complex accounting logic that derives from selling block space in advance and uh so yeah uh the design rational is there there have has been a lot of talk around this kind of stuff like inclusion pre confirmations uh and I mean splitting the markets for blog space but it it seems to us that a lot of The Proposal just want to go all in okay this seems like a more uh like we're taking it slow right uh so we're leaving the top partes um traditional Builders Meb boost Builders don't have to change anything except for the block size they're building for they're building building for a maximum size of of five million less but usually on average this is is not important uh that doesn't really matter to them uh the bottom part is then dedicated to a whole another market and it's allocated differently okay more efficiently uh and so for example uh it will I mean it would be suitable for all kind of position in sensitive transactions like uh blobs um like rewards distribution all kinds of like non um uh position sensitive U transactions you can think of uh the capacity can be sold ahead of time so I mean you can also in the Pres eventual presence of a secondary Market in which you can resell your gas tokens you bought in advance you can also do some kind of speculation on it or if you are a like a traditional M boost Builder that really needs some more space to fit their bundles in they can buy so I mean they're uh some some of the tokens on the secondary Market if they end up needing them like uh urgently and this basically introduces no breaking changes to the pro protocol this requires only like one relay like body ders to add one relay as a priority relay and uh in the future if we want to go relay agnostic basically some of the I mean some relays will new relays will spin up or some of the relays that are already existing will have to adapt a bit and run like X a specific software but this entails basically no changes at the protocol level uh and actually for this current version uh basically we don't need also any kind of changes on the like Meb boost software if you want we want to go relay agnostic yes we will minimal changes to MB boost uh and this allows for gradual changes and and experiment careful experimentation uh by just not going all in we're we're keeping the infrastructure as is for now and we can commit to more later okay uh so as I was as I was saying the bottom part of the block um will be probably uh also I mean gas tokens will be also probably uh sold on secondary markets uh which will be span up on the L2 XG L2 uh and we can like always adjust the parameters of of the markets uh both the like bottom part auction and the secondary Market parameters in order to like adapt to changes observations we can make uh about the market and we can probably imagine to um like splitting the Block in further more categories and uh yeah so let's go to conclusions so takeaways from this to blog space is not allocated efficiently uh the XGA okay is one proposal among several others to allocate block space more effic efficiently in a different way and split the markets uh our proposal allows for gradual change and can be extended over time uh and it can be integrated with other kinds of proposals so it's compatible with state state block auctions or with other proposals that are more uh like involved on the protocol level like multiple proposers uh or other distributed Builder mechanisms okay so I'm done with my talk and thank you all for attending if you have any questions yeah uh we have time maybe for one question let me get the mic if you can keep it short here you go uh thank you so my question is uh I don't get how uh using such Market I can still uh perform minor C value or Arbitrage because like essentially if I want to get on the top of a block my motivation is that there is either a transaction let's say Unis swap swap and I want to do a reverse swap before the Swap and then in the bottom I want to do re reverse Swap and like get back into into my position but then like if I only get to submit top part I first want to make conditional bottom and also I am targeting a particular transaction that I want to extract the value from and I don't see how this proposal addresses this this is I mean it doesn't need to address this it's like you're building a separate block that comes after that block you can do all the thing you say the things you've said in the bottom 25 million in the top 25 million gas so it's basically a block is 30 million Max okay we're just reserving some part which is going to be at the bottom but it doesn't matter if you want to back run your own transactions you can do that in the top part that makes sense thank you awesome I saw another hand before do you want to ask question question yeah go hi uh so doesn't the L2 add a new dependency and can this become a problem if you be become the main player uh in the game uh yeah totally uh I mean we can expect if this proposal is successful we can expect other services like that to spin up and then Market will differentiate so yeah that's the Hope okay awesome thank you very very much let's make a final Applause and we are now closing down the morning so please let's have a lunch at Palazo graina which is just nearby here and we will start back again at 2 thank you very much