the Hitchhikers Guide to Ethereum 2022（英文）行业资料

TheHitchhiker,sGuidetoEthereumJonCharbonneau EditorPick(Featured)MEMBERSOFOURTEAMHOLDETH.THISSTATEMENTISINTENDEDTODISCLOSEANYCONFLICTOFINTERESTANDSHOULDNOTBEMISCONSTRUEDASARECOMMENDATIONTOPURCHASEANYTOKEN.THISCONTENTISFORINFORMATIONALPURPOSESONLYANDYOUSHOULDNOTkeawaysEthereumistheonlymajorprotocolbuildingascalableunifiedsettlementAllroadsleadtotheendgameofcentralizedblockproduction,tlessblockvalidationandcensorshipresistanceInnovationssuchasproposer-builderseparationandweakstatelessnessunlockthisseparationofpowers(buildingandvalidating)toachieveDankshardingcombinesmultipleavenuesofcuttingedgeresearchtoprovidethescalablebaselayerrequiredforEthereum，srollup-centricofContentsg3)KZGCommitmentsKZGCommitmentsvsFraudProofs6)CensorshipResistanceList(crList)7)2-DimensionalKZGSchemenDankshardingKeyTakeawaysPimitEIP4)WeakStatelessnesseSecretLeaderElectionI，vebeenprettyskepticalonthetimingofthemergeeversinceVitaliksaidancepeopleborntodaylivetotheyearandhehopestobeimmortal.Butwhatthehell,let，shavesomefunandlookevenfurtherThisisnoquick-hitterarticle.Ifyouwantabroadyetnuancedunderstandingofoneoverarchinggoal一scalecomputationwithoutsacrificingdecentralizedsomecentralizationisneededtoscale.The“C”wordisscaryinblockchain,butit,strue.WejustneedtokeepthatpowerincheckwithdecentralizedandsforboththeLandaboveEthereumremainsincrediblysecurethrougheasydecentralizedvalidation,androllupsinherittheirsecurityfromtheL1.Ethereumthenprovidessettlementanddataavailabilityallowingrollupstoscale.Alloftheresearchhereultimatelylookstooptimizethesetworoleswhilesimultaneouslymakingiteasierthanevertofullyvalidatethechain.Hopefullyyou，veheardbynowthatEthereumhaspivotedtoarollup-centricapNomoreexecutionshardsEthereumwillinsteadoptimizefordatadatawasmadeavailable.andwhyDAisimportant.Ifyou，reunfamiliarorjustneedarefresher,Can，sntCelestiareportcoversthemShardProposersEachofthe64shardblockshadseparateproposersandcommitteesrotatingThisdesignintroducesunnecessarycomplexity,worseUX,andattackvectors.singleslotunlessyouintroduceverytightsynchronyassumptions.TheBeaconblockproposerneedstogatheralloftheindividualcommitteevotes,andtherecanbedelays.DSiscompletelydifferent.ValidatorsconductDAStoconfirmthatalldataisavailable(nomoreseparateshardcommittees).AspecializedbuildercreatesonelargeblockwiththeBeaconblockandallsharddataconfirmedtogether.PBSisthusnecessaryforDStoremaindecentralized(buildingthatlargeblocktogetherisresourceintensive).mplingallofit.Thatwouldmeanhighresourcerequirements,thushurtingalizationNaivesolution一Justcheckabunchofrandomchunksfromtheblock.IfartsolutionErasurecodethedatafirstExtendthedatausingaReedSolomoncode.Thismeansthedataisinterpolatedasapolynomial,andQuicklessonforeveryoneelsewhoforgotmathclass.(Ipromisethiswon，tbereallyscarymath一IhadtowatchsomeKhanAcademyvideostowritethesewPolynomialsareexpressionssumminganyfinitenumberoftermsoftheform.Thedegreeisthehighestexponent.Forexample,2x3+6x2+2x-4isapolynomialofdegreethree.Youcanreconstructanypolynomialofdegreedfromanyd+1coordinatesthatlieonthatpolynomial.Nowforaconcreteexample.Belowwehavefourchunksofdata(d0throughd3).Thesechunksofdatacanbemappedtoevaluationsofapolynomialf(X)atagivenpoint.Forexample,f(0)=d0.Nowyoufindthepolynomialofminimaldegreethatrunsthroughtheseevaluations.Sincethisisfourchunks,wecanfindthepolynomialofdegreethree.Then,wecanextendthisdatatoaddfourmoreevaluations(e0thoughe3)whichliealongthesamepolynomial.Rememberthatkeypolynomialproperty-wecanreconstructitfromanyfourpoints,notjustouroriginalfourdatachunks.BacktoourDAS.Nowweonlyneedtobesurethatany50%(4/8)oftheerasure-codeddataisavailable.Fromthat,wecanreconstructtheentireblock.Soanattackerwouldhavetohide>50%oftheblocktosuccessfullytrickDASnodesintothinkingthedatawasmadeavailablewhenitwasn,t.Theprobabilityof<50%beingavailableaftermanysuccessfulrandomsamplesisverysmall.Ifwesuccessfullysampledtheerasurecodeddata30times,thentheprobabilitythat<50%isavailableis2-30.KZGCommitmentsOk,sowedidabunchofrandomsamplesanditwasallavailable.Butwehaveanotherproblem-wasthedataerasurecodedproperly?Otherwisemaybetheblockproducerjustadded50%junkwhentheyextendedtheblock,andwesamplednonsense.Inwhichcasewewouldn,tactuallybeabletoreconstructthedata.However,wealsoneedtoknowthatalloftheoriginalandextendeddatalieony1.Celestiaisgoingthefraudproofroute.Somebodyneedstowatch,andiftheblockiserasurecodedincorrectlythey,llsubmitafraudprooftoalerteveryone.Thisrequiresthestandardhonestminorityassumptionandsynchronyassumption(i.e.,inadditiontosomeonesendingmethefraudndwillreceiveitwithinafinite2.EthereumandPolygonAvailaregoinganewroute一KZGcommitmentsassumptionsforsafetyinregardstofraudproofsthoughthey,restillthereOthersolutionsexist,butthey,renotbeingactivelypursued.Forexample,youHowever,they,reexpectedtoimproveoverthenextfewyears,soEthereumwilllikelypivottoSTARKsdowntheroadbecauseKZGcommitmentsarenotquantum-resistant.BacktoKZGcommitments-theseareatypeofpolynomialcommitmentscheme.Commitmentschemesarejustacryptographicwaytoprovablycommittosomevalues.Thebestmetaphorisputtingaletterinsidealockedboxandhandingittosomeoneelse.Thelettercan,tbechangedonceinside,butitcanbeopenedwiththekeyandproven.Youcommittotheletter,andthekeyistheproof.Inourcase,wemapallthatoriginalandextendeddataonanX,Ygridthenfindtheminimaldegreepolynomialthatrunsthroughthem(thisprocessiscalledaLagrangeinterpolation).Thispolynomialiswhattheproverwillcommitto:Herearethekeypoints:Wehavea“polynomial”f(X)Theprovermakesa“commitment”C(f)tothepolynomialThisreliesonellipticcurvecryptographywithatrustedsetup.Forabitmoredetailonhowthisworks,here,sagreatthreadfromBartekForany“evaluation”y=f(z)ofthispolynomial,theprovercancomputea“proof”(f,z)GiventhecommitmentC(f),theproof(f,z),anypositionz,andtheevaluationyofthepolynomialatz,averifiercanconfirmthatindeedf(z)=yTranslation:theprovergivesthesepiecestoanyverifier,thentheverifiercanconfirmthattheevaluationofsomepoint(wheretheevaluationrepresentstheunderlyingdata)correctlyliesonthepolynomialthatwascommittedtoThisprovesthattheoriginaldatawasextendedcorrectlybecauseallevaluationslieonthesamepolynomialNoticetheverifierdoesn,tneedthepolynomialf(X)Importantproperty-thishasO(1)commitmentsize,O(1)proofsize,andO(1)verificationtime.Evenfortheprover,commitmentandproofgenerationonlyscaleatO(d)wheredisthedegreeofthepolynomialTranslation:evenasn(thenumberofvaluesinX)increases(i.e.,asthedatasetincreaseswithlargershardblobs)-thecommitmentsandproofsstayconstantsize,andverificationtakesaconstantamountofeffortThecommitmentC(f)andproof(f,z)arebothjustoneellipticcurveelementonapairingfriendlycurve(thiswilluseBLS12-381).Inthiscase,theywouldbeonly48byteseach(reallysmall)Soaprovercommitstoahugeamountoforiginalandextendeddata(representedasmanyevaluationsalongthepolynomial)atstilljust48bytes,andtheproofwillalsojustbe48bytesTLDR-thisishighlyscalableTheKZGroot(apolynomialcommitment)isthenanalogoustoaMerkleroot(whichisavectorcommitment):Theoriginaldataisthepolynomialf(X)evaluatedatthepositionsf(0)tof(3),thenweextenditbyevaluatingthepolynomialatf(4)tof(7).Allpointsf(0)tof(7)areguaranteedtobeonthesamepolynomial.Bottomline:DASallowsustocheckthaterasurecodeddatawasmadeavailable.KZGcommitmentsprovetousthattheoriginaldatawasextendedTakeastepbacktocomparethetwoapproachesnowthatweunderstandhowKZGworks.veandthetrustedsetup(whichisopentoparticipatein)requiresonlyasinglehonestcipantKZGadvantages-lowerlatencythanthefraudproofsetup(thoughasnotedGasperwon，thavefastfinalityanyway),anditensurespropererasurecodingwithoutintroducingthesynchronyandhonestminorityassumptionsinherentinHowever,considerthatEthereumwillstillreintroducetheseassumptionsforblockreconstruction,soyou，renotactuallyremovingthem.DAlayersalwaysneedtoplanforthescenariowheretheblockwasmadeavailableinitially,butneedtocommunicatewitheachothertopieceitbacktogetherThis1.Youhaveenoughnodes(lightorfull)samplingthedatasuchthattopieceitbacktogetherThisisafairlyweakandunavoidablehonestminorityassumption,sonotahugeconcern.2.Reintroducesthesynchronyassumption-thenodesneedtobeabletocommunicatewithinsomeperiodoftimetoputitbacktogether.EthereumvalidatorsfullydownloadshardblobsinPDS,andwithDSthey,llonlyconductDAS(downloadingassignedrowsandcolumns).Celestiawillrequirevalidatorstodownloadtheentireblock.Notethatineithercase,weneedthesynchronyassumptionforreconstruction.Intheeventtheblockisonlypartiallyavailable,fullnodesmustcommunicatewithothernodestopieceitbacktogether.ThelatencyadvantageofKZGwouldshowupifCelestiaeverwantedtomovefromrequiringvalidatorstodownloadtheentiredatatoonlyconductingDAS(thoughthisshiftisn,tcurrentlyplanned).TheywouldthenneedtoimplementKZGcommitmentsaswell-waitingforfraudproofswouldmeansignificantlyincreasingtheblockintervals,andthedangerofvalidatorsvotingforanincorrectlyencodedblockwouldbemuchtoohigh.IrecommendthebelowfordeeperexplorationofhowKZGcommitmentswork:A(RelativelyEasyToUnderstand)PrimeronEllipticCurveCryptographyExploringEllipticCurvePairings-VitalikKZGpolynomialcommitments-DankradHowdotrustedsetupswork?-VitalikIn-protocolProposer-BuilderSeparationConsensusnodestoday(miners)andafterthemerge(validators)servetworoles.Theybuildtheactualblock,thentheyproposeittootherconsensusnodeswhovalidateit.Miners“vote”bybuildingontopofthepreviousblock,andafterthemergevalidatorswillvotedirectlyonblocksasvalidorinvalid.PBSsplitstheseup-itexplicitlycreatesanewin-protocolbuilderrole.Specializedbuilderswillputtogetherblocksandbidforproposers(validators)toselecttheirblock.ThiscombatsMEV,scentralizingforce.RecallVitalik,s“Endgame”-allroadsleadtocentralizedblockproductionwithtrustlessanddecentralizedvalidation.PBScodifiesthis.WeneedonehonestBuildersreceivepriorityfeetipspluswhateverMEVtheycanextract.Inanefficientmarket,competitivebuilderswillthenbiduptothefullvaluetheycanextractfromblocks(lesstheiramortizedcostssuchaspowerfulhardware,etc.).TheexactPBSimplementationisstillbeingdiscussed,buttwo-slotPBScouldids2.Beaconblockproposerchoosesthewinningheaderandbid.ProposerisProposersareselectedfromthevalidatorsetusingthestandardRANDAOmechanism.Thenweuseacommit-revealschemewherethefullblockbodyThecommit-revealismoreefficient(sendingaroundhundredsoffullblockbodiescouldoverwhelmbandwidthonthep2player),anditalsopreventsMEVstealing.Ifbuildersweretosubmittheirfullblock,anotherbuildercouldseeit,figureoutthatstrategy,incorporateit,andquicklypublishabetterblock.Additionally,sophisticatedproposerscoulddetecttheMEVstrategyusedandcopyitwithoutcompensatingthebuilder.IfthisMEVstealingbecametheequilibriumitwouldincentivizemergingthebuilderandproposer,soweavoidAfterthewinningblockheaderhasbeenselectedbytheproposer,thecommitteeconfirmsitandsolidifiesitintheforkchoicerule.Then,thewinningbuilderpublishestheirwinningfull“builderblock”body.Ifitwaspublishedinthefullbidtotheproposer(andloseoutonalloftheMEVandfees).Thisderfixed12seconds,soherewe，dneed24secondsforafullblocktime(two12-otsifwedontwanttointroduceanynewassumptionsGoingwithsecondsforeachslotsecondblocktimes)appearstobeasafecompromise,sogoodattheirjobthatalltheotherbuildersgaveup,ormaybethey，lljustucrListsputacheckonthispower.Theexactimplementationisagainanopendesignspace,but“hybridPBS”appearstobethefavorite.Proposersspecifyalistofalleligibletransactionstheyseeinthemempool,andthebuilderwillbeforcedtoincludethem(unlesstheblockisfull):1.ProposerpublishesacrListandcrListsummarywhichincludesalleligibletransactions2.BuildercreatesaproposedblockbodythensubmitsabidwhichincludesahashofthecrListsummaryprovingthey,veseenit3.Proposeracceptsthewinningbuilder,sbidandblockheader(theydon,tuilderpublishestheirblockandincludesaprooftheyveincludedalltransactionsfromthecrListorthattheblockwasfull.OtherwisetheblockTherearestillimportantquestionstobeironedouthere.Forexample,thedominanteconomicstrategyhereisforaproposertosubmitanemptylist.Thiswouldallowevencensoringbuilderstowintheauctionsolongastheybidthe2-DimensionalKZGSchemeWesawhowKZGcommitmentsallowustocommittodataandproveitwascommittoallofthedatainasingleKZGcommitment-asingleblockwillusemanyKZGcommitments.Wealreadyhaveaspecializedbuilder,sowhynotjusthavethemcreateonegiantKZGcommitment?Theissueisthatthiswouldrequireapowerfulsupernodetoreconstruct.We，reokwithasupernoderequirementforinitialmanyKZGcommitmentsmakesthisfeasible.Reconstructionmayevenbefairlyountofdataathand.ereachblockwillincludemshardblobsencodedinInstead,Ethereumwillusea2DKZGscheme.WeusetheReed-SolomoncodeWemakeita2DschemebyextendingadditionalKZGcommitments(256-511here)thatlieonthesamepolynomialas0-255.NowwejustconductDASonthetableabovetoensureavailabilityofdataacrossallshards.The2Dsamplingrequirementfor75%ofthedatatobeavailable(asopposedto50%earlier)meanswedoabithighernumberoffixedsamples.BeforeImentioned30samplesforDASinthesimple1Dscheme,butthiswillrequire75samplestoensurethesameprobabilisticoddsofreconstructinganavailableblock.Sharding1.0(whichhada1DKZGcommitmentscheme)onlyrequired30oundwithouteverindividuallycheckingallshards.Thecombinedblockswiththe2DKZGcommitmentschemenowmakesittrivialset.However,DankradrecentlytookadvantageofthatdesignrealizingthatitChainexecutionblockandshards.WenowhaveonebuildercreatingtheentirebeinfeasiblewithoutPBS-regularvalidatorscouldn，thandlethemassivesSharding1.0included64separatecommitteesandproposers,soeachshardcouldhavebeenunavailableindividually.ThetighterintegrationhereallowsustoensureDAinaggregateinoneshot.Thedataisstill“sharded”underthehood,butfromapracticalperspectivedankshardingstartstofeelmorelikebigblockswhichisgreat.Danksharding-HonestMajorityValidationValidatorsattesttodatabeingavailableasfollows:Thisreliesonthehonestmajorityofvalidators-asanindividualvalidator,mythatthefullblockisavailable.Thisreliesonthehonestmajoritytosayitis.tionmattersNotethisisdifferentfromthe75randomsampleswediscussedearlier.Privaterandomsamplingishowlow-resourceindividualswillbeabletoeasilycheckavailability(e.g.,IcanrunaDASlightnodeandknowtheblockisavailable).However,validatorswillcontinuetousetherowsandcolumnsapproachfoructionAslongas50%ofanindividualroworcolumnisavailable,thenitcaneasilybefullyreconstructedbythesamplingvalidator.Astheyreconstructanychunkslines.ThishelpsothervalidatorsreconstructanymissingchunksfromtheirowsandcolumnsasneededEnoughnodestoperformsamplerequestssuchthattheycollectivelyhavekSynchronyassumptionbetweennodeswhoarebroadcastingtheirSo,howmanynodesisenough?Roughestimatesballparkit~64,000individualinstances(currentlyover~380,000todaysofar).Thisisalsoaverypessimisticcalculationwhichassumesnocrossoverinnodesbeingrunbythesamevalidator(whichisfarfromthecaseasnodesarelimitedto32ETHinstances).Ifcancollectivelyretrievethembecauseofthecrossover.ThisstartstoscaleencomfortablylowDScanbesetaticallyThusthesecurityassumptionDanksharding-MaliciousMajoritySafetyWithPrivateRandomSamplingindividualcan，tprovetomyselfthatablockisavailablebyjustdownloadingacouplerowsandcolumns.However,privaterandomsamplingcangivemethisguaranteewithouttrustinganyone.Thisiswherenodescheckthose75randomsamplesasdiscussedearlier.DSinitiallywon，tincludeprivaterandomsamplingbecauseit，ssimplyaveryrnyoutheexactchunksyourequestedandwithholdtherest.Soyouwouldn，tknowDanksharding-KeyTakeawaysInadditiontobeingasweetname,DSisalsoincrediblyexciting.ItfinallyachievesEthereum，svisionofaunifiedsettlementandDAlayer.ThistightInfact,let,sdefinewhyit,sevenconsidered“sharded”atall.Theonlyremnantof“sharding”issimplythefactthatvalidatorsaren,tresponsiblefordownloadingallofthedata.That,sit.Soyou,renotcrazyifyouwerequestioningbynowwhetherthisisreallystillsharding.ThisdistinctioniswhyPDS(we,llcoverthisshortly)isnotconsidered“sharded”(eventhoughithas“sharding”inthename,yesIknowit,sconfusing).PDSrequireseachvalidatortofullydownloadallshardblobsinordertoattesttotheiravailability.DSthenintroducessampling,soindividualvalidatorsonlydownloadpiecesofit.Theminimalshardingthankfullymeansamuchsimplerdesignthansharding1.0(soquickerdeliveryright?right?).Simplificationsinclude:LikelyhundredsoflineslesscodeintheDSspecificationvs.thesharding1.0specification(thousandsoflineslessintheclients)Nomoreshardcommitteeinfrastructure,committeesjustneedtovoteonthemainchainNotrackingofseparateshardblobconfirmations,nowtheyallgetTheremovalofshardcommitteesalsostrengthensbriberyresistance.DSvalidatorsvoteonceperepochontheentireblock,sodatagetsconfirmedby1/32oftheentirevalidatorsetimmediately(32slotsperepoch).Sharding1.0validatorsalsovotedonceperepoch,buteachshardhaditsowncommitteesThecombinedblockswiththe2DKZGcommitmentschemealsomakesDASfarmoreefficientasdiscussed.Sharding1.0wouldrequire60KB/sbandwidthtoAnotherexcitingpossibilitykeptalivewithDS一synchronouscallsbetweenZK-rollupsandL1Ethereumexecution.TransactionsfromashardblobcanimmediatelyconfirmandwritetotheL1becauseeverythingisproducedinthenChainblockShardingwouldveremovedthispossibilityduetoseparateshardconfirmations.ThisallowsforanexcitingdesignspacewhichityweseetodayAddingmorenodestoaTherearestilllimitstoblockchainscalability,butwecanpushordersofmagnitudehigherthananythingweseetoday.Secureandscalablebaselayersallowexecutiontoproliferateatopthem.ImprovementsindatastorageandPushingbeyondtheDAthroughputcontemplatedhereiscertainlyinthecards,butit,shardtosaywherethatmaxwillendup.Thereisn,taclearredline,butroleisn,ttoguaranteedataretrievabilityindefinitely.Itsroleistomakeitavailableforenoughtimesuchthatanyonewhocarestodownloaditcan,satisfyingoursecurityassumptions.Thenitgetsdumpedintostoragewherever一thisiscomfortableashistoryisa1ofNtrustassumption,andwe,renotactuallytalkingaboutthatmuchdatainthegrandschemeofthings.ThiscouldgetintouncomfortableterritorythoughyearsdowntheValidators一DASrequiresenoughnodestocollectivelyreconstructtheblock.Otherwise,anattackercouldwaitaroundandonlyrespondtotherovidedarentenoughtoreconstructtheblock,theattackercouldwithholdtherestandwe,reoutofluck.Tosafelyincreasethroughput,weneedtoaddmoreDASnodesorincreasetheirdatabandwidthrequirements.Thisisn,taconcernforthethroughputdiscussedhere.Againthough,thiscouldgetuncomfortableifthroughputNoticethebuilderisn,tthebottleneck.You,llneedtoquicklygenerateKZGproofsfor32MBofdata,soexpectaGPUorprettybeefyCPUplusatleast2.5GBit/sbandwidth.ThisisaspecializedroleanywayforwhomthisisanegligibleProto-danksharding(EIP-4844)DSisawesome,butwe,llhavetobepatient.PDSaimstotideusover一itimplementsnecessaryforward-compatiblestepstowardDSonanexpeditedintheinterim.However,itdoesn,tactuallyimplementdatashardingyet(i.e.,RollupstodayuseL1“calldata”forstoragewhichpersistson-chainforever.However,rollupsonlyneedDAforsomereasonableperiodoftimesuchthatEIP-4844introducesthenewblob-carryingtransactionformatthatrollupswillandtheycanbemuchcheaperthansimilaramountsofcalldata.DatablobsareForcontextonscale,currentEthereumblocksaregenerallyaverage~90KB(calldatais~10KBofthis).PDSunlocksfarmoreDAbandwidth(target~1MBAblobisavectorof4096field-elementsof32byteseach.PDSallowsforamaxOrdersofmagnitudescalingwitheachstep.PDSstillrequiresconsensusnodeseconservativeDSdistributestheloadofKZGcommitmentstotheblobsSLayerseparationbetweenBeaconBlockverificationandDASblobsePBSDAS2DKZGschemeProof-of-custodyorsimilarin-protocolrequirementforeachvalidatortoverifyavailabilityofaparticularpartoftheshardeddataineachblockEVMonlyviewsthecommitmentattachedtotheblobs.TheexecutionlayermadewithEIParealsoforwardcompatiblewithDSandnomorealterationswillbeneededonthisside.TheupgradefromPDStoDSthenonlyrequiresconsensuslayerchanges.byconsensusclientsinPDSTheblobsarenowreferenced,butnotfullyencoded,intheBeaconblockbody.Insteadofembeddingthefullcontentsinthebody,thecontentsoftheblobsareWediscussedearlierhowtocommittotheblobsusingKZGpolynomialcommitments.However,insteadofusingtheKZGdirectly,EIP-4844implementsEVMcompatibility-KZGcommitmentsare48byteswhereastheEVMForwardcompatibility-ifweeverswitchfromKZGtosomethingelse(STARKsforquantum-resistance),thecommitmentscancontinuetobe32MultidimensionalEIP-1559marketwithseparatefloatinggaspricesandlimits.SoevenifsomeNFTprojectissellingabunchofmonkeylandonL1,yourrollupdatacostswon，tgoup(thoughproofsettlementcostswould).ThisacknowledgesthatthedominantfstsownEIP-1559mechanism.ThelongrunaveragenumberofblobsperblockYoueffectivelyhavetwoauctionsrunninginparallel-oneforcomputationandngTherearesomeinterestingdesignsbeingthrownaroundhere.Forexample,itmightmakesensetochangeboththecurrentgasandblobpricingmechanismsfromthelinearEIP-1559toanewexponentialEIP-1559mechanism.Thecurrentimplementationdoesn，taverageouttoourtargetblocksizesinpractice.ThetabilizesimperfectlytodayresultingintheobservedaveragegasusedPartIIHistory&StateManagementQuickrecaponsomebasicshere:tickitonState一Snapshotofallthecurrentaccountbalances,smartcontracts,etc.Fullnodes(currently)allneedthisonhandtovalidatetransactions.It,stoobigforRAM,andaharddriveistooslow一itgoesinyourSSD.Highthroughputblockchainsballoontheirstate,growingfarbeyondwhatusnormiescankeeponourlaptops.Ifeverydayuserscan,tholdthestate,onTLDR一thesethingsgetreallybig,soifyoumakenodesholdontothemitgetshatsCalldataGasCostReductionWithTotalCalldataLimit(EIP-4488)PDSisagreatsteppingstonetowardDSwhichchecksoffmanyoftheeventualrequirements.ImplementingPDSwithinareasonabletimespancanthenpullitaddressesthefeeemergencynonethelessUnfortunatelyitdoesn,timplementstepsalongthewaytoDS,soalloftheinevitablechangeswillst