ChromiaPlatformwhitepaper平臺(tái)白皮Platformwhitepaper平臺(tái)白皮書(shū)ExecutiveSummary摘要Motivation動(dòng)Technicaldesign&features技術(shù)設(shè)計(jì)&特征Uses使用Designrationale設(shè)計(jì)原Overviewofproblemswithexistingplatforms現(xiàn)有平臺(tái)問(wèn)題概述Blockchainasadatabase作為數(shù)據(jù)庫(kù)的區(qū)塊鏈Relationalmodel關(guān)系模First--classdecentralizedapplications頂尖的分散應(yīng)用程序Programmingmodel編程模型Consensus&nodes共識(shí)&節(jié)點(diǎn)Modeloverview模型概述SybilcontrolmechanismSybil控制機(jī)制Consensus共識(shí)Nodecompensation節(jié)點(diǎn)補(bǔ)償Miscellaneousfeatures其他特征Decentralizedapplications分散應(yīng)用程序Transparentapps透明的應(yīng)用程序Tokenmodel代幣模型TheroleofChromiaChromia的角Notcontrolledbyasingleentity.非受控于單個(gè)個(gè)體Controlledbythecommunityofusers.受用戶(hù)社區(qū)控制Cannotbeshutdown.不可關(guān)閉Censorship--resistant.抗審查T(mén)ransparent.透明的Privacy.隱私Highlyavailable.高效Decentralizationquality去中心化質(zhì)量Platformarchitecture平臺(tái)架構(gòu)Postchain后鏈ChainsNodeimplementation節(jié)點(diǎn)實(shí)Interactionwithotherblockchains與其他區(qū)塊鏈的交互Components組件Governance管ChromapolissystemgovernanceChromia系統(tǒng)管Initialcentralization初始中心化Rejectedalternatives被否決的選擇Stake/coinvoting賭注、錢(qián)幣投Noformalgovernance沒(méi)有正式的管理Uniqueusers獨(dú)特的用戶(hù)Applicationgovernance應(yīng)用程序管理Uses使用Tokens代幣Games游Businessuses商業(yè)用途Tokensandincentives代幣和激勵(lì)Fees費(fèi)Applicationfeemodels應(yīng)用程序費(fèi)用模型Hostingfees代管費(fèi)Nodeincentives節(jié)點(diǎn)激勵(lì)Nodestakes節(jié)點(diǎn)股份Tokenuseingames游戲中使用的代幣Chromatokeneconomics幻彩代幣經(jīng)濟(jì)Systemaccounts系統(tǒng)賬戶(hù)Publicgoodaccount公共優(yōu)良賬戶(hù)Tokendistribution代幣分配Promotionaltokenfund促銷(xiāo)代幣資金Decentralization去中心化Centralizationnecessaryatstart以中心化開(kāi)Decentralizationthroughadiversesetofproviders通過(guò)多樣化的供應(yīng)商去中心化Bitcoin比特幣DPoSEthereum以太坊ChromapolisChromiaNumberoffullnodes全節(jié)點(diǎn)數(shù)量Security安全Blockchain區(qū)塊鏈Nodesecurity節(jié)點(diǎn)安全Governancesecurity管理安全Lightclientsecurity輕客戶(hù)安全Dappclientandwalletsecurity分散應(yīng)用程序客戶(hù)和錢(qián)包安ExecutiveSummary摘Chromiaisanewblockchainplatformfordecentralizedapplications,conceivedinresponsetotheshortcomingsofexistingplatformsanddesignedtoenableanewgenerationofdappstoscalebeyondwhatiscurrentlypossible.Chromia是為分散應(yīng)用程序提供的一個(gè)嶄新的區(qū)塊鏈平臺(tái)，旨在解決現(xiàn)存平臺(tái)的缺陷，開(kāi)創(chuàng)分散應(yīng)用程序的新時(shí)代。Motivation動(dòng)WhileplatformssuchasEthereumallowanykindofapplicationtobeimplementedintheory,inpracticetheyhavemanylimitations:baduserexperience,highfees,frustratingdeveloperexperience,poorsecurity.Thispreventsdecentralizedapps(dapps)fromgoingmainstream.雖然像以太坊這樣的平臺(tái)理論上可支持任何應(yīng)用程序的執(zhí)行，但實(shí)際運(yùn)用中卻存在很多局限：不良的用戶(hù)體驗(yàn)、高昂的平臺(tái)費(fèi)用、欠佳的開(kāi)發(fā)者體驗(yàn)、較低的安全性。這些都阻礙了分散應(yīng)用程序走向主流的步伐。Webelievethattoaddresstheseproblemsproperlyweneedtoseriouslyrethinktheblock--chainarchitectureandprogrammingmodelwiththeneedsofdecentralizedapplicationsinmind.Ourprioritiesareto:要解決這一系列問(wèn)題，我們需要重新認(rèn)真了解區(qū)塊鏈架構(gòu)，以分散應(yīng)用程序的需求為中心來(lái)設(shè)計(jì)模塊。我們的首要任務(wù)是：Allowdappstoscaletomillionsofusers.使分散應(yīng)用程序面向數(shù)百萬(wàn)用戶(hù)Maketheuserexperienceassmoothaswithacentralizedapplication:nofeeforeveryinteraction,nowaitingtime.使用戶(hù)體驗(yàn)像中心化應(yīng)用程序一樣平穩(wěn)，每筆交互都不收Allowdeveloperstobuildsecureapplicationswithminimaleffort,usingfamiliarpara--digms.使開(kāi)發(fā)者能用最少的精力，參照相似的范例來(lái)開(kāi)發(fā)安全的應(yīng)用程序。Technicaldesign&features技術(shù)設(shè)計(jì)和特Webelievethatwithinadecentralizedapplicationecosystemablockchainservestheroleofashareddatabase:Itstoresapplicationdataandmakessurethatdataadditions,updatesandtransformationsareauthorizedandconsistentwiththeapplication’srules.Forthisreason,Chromiaisdesignedandoptimizedtoservetheroleofashareddatabaseinthebestwaypossible.Itfeatures:在一個(gè)分散應(yīng)用程序的生態(tài)系統(tǒng)中，區(qū)塊鏈扮演著共享數(shù)據(jù)庫(kù)的角色。它存儲(chǔ)應(yīng)用程序數(shù)據(jù)確保數(shù)據(jù)的添加、更新、轉(zhuǎn)化都有授權(quán)，且遵照應(yīng)用程序的規(guī)則。為此，Chromia應(yīng)運(yùn)而生，力爭(zhēng)成為最佳的共享數(shù)據(jù)庫(kù)。它的特征有：Arelationalmodel1:Blockchaindataandapplicationstatearestoredinarelationaldatabase.Thismodelisconsideredtobebestinclassintermsofflexibility,versatilityandconsistency.關(guān)系模型：區(qū)塊鏈數(shù)據(jù)和應(yīng)用程序狀態(tài)被存儲(chǔ)在關(guān)系數(shù)據(jù)庫(kù)中。這一模塊被認(rèn)為在靈活性、通用性和相容性方面是最一流的。Horizontalscaling:Eachdappgetsitsownblockchain(or,perhaps,multipleblock--chains).Eachblockchainisrunbyasubsetofnodes,thusbyincreasingnumberofnodeswecanincreasetotalthroughput.橫向擴(kuò)展：每一個(gè)分散應(yīng)用程序擁有自己的區(qū)塊鏈（或許可能有多個(gè)區(qū)塊鏈）。每一區(qū)塊鏈通過(guò)節(jié)點(diǎn)的子集來(lái)運(yùn)行，這樣就可以通過(guò)增加節(jié)點(diǎn)數(shù)量來(lái)增加總產(chǎn)量。Richindexingandquerying:Dappscanquicklyretrieveinformationtheyneeddirectlyfromnodesrunningtheapplication.Dappblockchainlogiccanperformcomplexque--rieswithoutsevereperformancedegradation.強(qiáng)大的檢索和查詢(xún)：分散應(yīng)用程序能快速?gòu)倪\(yùn)行的應(yīng)用程序節(jié)點(diǎn)中直接獲取它們需要的信息。分散應(yīng)用程序的區(qū)塊鏈邏輯可以在沒(méi)有嚴(yán)重績(jī)效降級(jí)的情況下完成復(fù)雜的查詢(xún)。Arelationalprogramminglanguage:Chromiadappbackendsarewritteninaspecial--izedlanguagewhichisdeeplyintegratedwiththerelationalmodel.Thismodelin--creasesprogrammerproductivityandensuresapplicationconsistency.關(guān)系編程語(yǔ)言：Chromia的分散應(yīng)用程序后端由專(zhuān)業(yè)語(yǔ)言編成，與關(guān)系模型深入融HighI/Othroughput:dataqueriesandupdatesaredelegatedtoaheavilyoptimizedrelationaldatabase,allowingdappstoperformalargenumberofqueriesanddataupdateoperations.I/O高吞吐量：數(shù)據(jù)查詢(xún)和更新的速度代表了極度優(yōu)化的關(guān)系數(shù)據(jù)庫(kù)，使分散應(yīng)用程序能處理大量查詢(xún)和數(shù)據(jù)更新。PBFT2--styleconsensus:Transactionscanbeconfirmedwithinseconds.PBFT式共識(shí)：交易在數(shù)秒內(nèi)就能確認(rèn)。First--classdapps:Dappsdonotarisefrom“smartcontracts”inChromia,butarecon--sideredfirst--classentities.Chromiagivesdappdevelopersahighdegreeofflexibilityandcontrol.Forexample,Chromiadoesnotchargedappusersforeachtransactiontheymake,insteadcollectingfeesfromdappsasawhole.Thisleavesdevelopersfreetocreatetheirownfeeandresourceusepolicies.一流的分散應(yīng)用程序：在Chromia，分散應(yīng)用程序并非起源于“智能合約”，而是一個(gè)一流的個(gè)體。Chromia給予分散應(yīng)用程序開(kāi)發(fā)者更強(qiáng)的靈活性和控制權(quán)。例如，Chromia不對(duì)分散用戶(hù)的每筆交易進(jìn)行收費(fèi)，而是整體收費(fèi)。這使得開(kāi)發(fā)者能自主制定費(fèi)用和資源使用政策。1CoddE.F(1970)ARelationalModelofDataforLargeSharedDataBanks"大型共享數(shù)據(jù)庫(kù)的數(shù)據(jù)關(guān)系模CommunicationsoftheACM.計(jì)算機(jī)協(xié)會(huì)通Classics13(6)377–87;/ci--2Castro,M.;?Liskov,B.(2002)."PracticalByzantineFaultToleranceandProactiveRecovery"“實(shí)用的拜占庭式容錯(cuò)和主動(dòng)恢復(fù)”.ACMTransactionsonComputerSystems.計(jì)算機(jī)系統(tǒng)上的ACM交易Asso--ciationforComputingMachinery.計(jì)算機(jī)協(xié)會(huì) 20(4):398–461.ChromiaisimplementedontopoftheexistingPostchain3frameworkdevelopedbyChro--Chromia在幻彩大道開(kāi)發(fā)的后鏈架構(gòu)上運(yùn)行Chromiaoffersthesamelevelofopenness,transparencyanddecentralizationasotherpublicblockchains.InChromiaminersarereplacedwithproviders.Providersownnodeswhichpro--duceblocks.IthasbeensuggestedthatthefourlargestminingpoolsofbothBitcoin4andEthereum5couldexertsignificantcontroloverthosenetworksiftheycolluded.Weaimtoen--surethattheminimumnumberofnodeproviderswhosecollusionwouldberequiredtoexertsuchcontrolonChromiaexceedsthisnumbersignificantly.ItcanthereforebesaidthattheChromiamodeldoesnottendtowardscentralizationanymorethantheoldestandmosttrustedpublicblockchains.Chromia與其他公有區(qū)塊鏈一樣開(kāi)放、透明、去中心化。在Chromia，礦工被供應(yīng)商所取代供應(yīng)商擁有能夠生產(chǎn)區(qū)塊的節(jié)點(diǎn)。有人建議，如果比特幣和以太坊的四大礦池串通，將對(duì)那些網(wǎng)絡(luò)形成極強(qiáng)的控制。而我們力求確保節(jié)點(diǎn)供應(yīng)商的最小量遠(yuǎn)超這一數(shù)字。那么我們可以說(shuō)Chromia模型并不像最古老的，最受信任的公有區(qū)塊鏈那樣走向中心化。Chromia’PBFT--styleconsensusisfurtherhardenedbyanchoring6tomakesurethatfinalityisatleastasstrongasthefinalityofProofofWork(PoW)blockchainssuchasBitcoinandEthereum.ToalterthehistoryofananchoredportionofChromiablockhistoryitwouldbenecessarytocombinePoWblockchainreorganizationwithamaliciouscollusionofasufficientnumberofChromianodes.ChromiaPBFT式共識(shí)通過(guò)側(cè)鏈來(lái)加固，確保它的終端至少同比特幣、以太坊這樣的PoW區(qū)鏈一樣強(qiáng)健。為了改變以往Chromia區(qū)塊中的側(cè)鏈比例，我們有必要將PoW區(qū)塊鏈與Chromia一定數(shù)量的惡意串通節(jié)點(diǎn)進(jìn)行重組Uses使Chromiaisageneral--purposeplatformwhichissuitableforalmostallkindsofdapps.ItisparticularlywellsuitedtocaseswhichrequirehighI/Ocapacityorwhichinvolvemanagementofcomplexdatasets.Chromia是一個(gè)面向大眾的平臺(tái)，它適用于幾乎所有種類(lèi)的分散應(yīng)用程序。尤其適用于對(duì)吞吐量有高要求的或者需要管理復(fù)雜數(shù)據(jù)的情況3456WeoriginallydescribedAnchoringas“side--chains”我們最初將錨定形容為“側(cè)鏈”/index.php?topic=313347;?amoreformaldiscussionofanchoringcanbefoundintheBitFurywhitepaper“OnBlockchainAuditability”.對(duì)于錨定更正式的討論可以在BitFury白皮書(shū)“區(qū) Massivelymultiplayeronlinegames(MMOGs)areoneexample.Chromiaiscapableofhost--ingentiregameworldsintheblockchain,makingsurethattheyevolveaccordingtopredeter--minedrulesandensuringthatnoonecancheat.Blockchaingamingisbecomingincreasinglypopular,butMMOGsarecurrentlyoutofreachbecausenoexistingblockchainplatformcansupportthem.WebelievethatimplementingaMMOGwillbethebestwaytoshowcasethecapabilitiesofChromia.MMOGshaveaverydemandingsetofrequirements;?thecapacitytorunMMOGsimpliesthatChromiaissuitablefordemandingandcomplexdappsofallkinds.多用戶(hù)在線(xiàn)游戲（MMOG）就是一個(gè)很好的例子。Chromia可在區(qū)塊鏈上支持整個(gè)游戲世界，并確保按照既定的規(guī)則進(jìn)行，沒(méi)有人可以作弊。區(qū)塊鏈游戲正日趨受歡迎，但MMOG卻很少有人涉足，因?yàn)楝F(xiàn)有的區(qū)塊鏈平臺(tái)都無(wú)法支持他們。而MMOG的運(yùn)行恰恰最好展示了Chromia的實(shí)力。MMOG對(duì)平臺(tái)有著十分苛刻的要求，因此MMOG的運(yùn)行意味著Chromia有能力應(yīng)對(duì)任何復(fù)雜的分散應(yīng)用程序。Designrationale設(shè)計(jì)原Overviewofproblemswithexistingplatforms現(xiàn)有平臺(tái)問(wèn)題概Ethereumwasafirstblockchaintoofferaplatformfordecentralizedapplicationdevelopment.Manyapplicationprototypeswerecreated,butdevelopersfacedthefollowingissues:以太坊是為分散應(yīng)用程序發(fā)展提供的第一個(gè)區(qū)塊鏈平臺(tái)。它創(chuàng)建了許多應(yīng)用程序標(biāo)準(zhǔn)，但開(kāi)發(fā)者會(huì)面臨以下問(wèn)題：Limitedcapacity.Becausenetworkcapacityislimited,andusagefeesareproportionaltoload,transactionfeescanbe$1ormoreforcomplexapplications.Thiscost,typi--callypaidforeachinteractionwithanapplication,makesmostapplicationstooexpen--sivetobepractical.有限容量。因?yàn)榫W(wǎng)絡(luò)容量有限，且使用費(fèi)是按比例收取，交易費(fèi)可達(dá)1美元，一些雜的應(yīng)用程序會(huì)更貴。高昂的成本使得大多數(shù)應(yīng)用程序都貴的離譜ProhibitivelyexpensiveI/Ooperations,forthesamereason.Forexample,acontractcannotiteratethroughalistofuserssincethecostofthisactionwouldexceedtheblockgaslimit.Thusdevelopershavetojumpthroughhoopstoimplementsomethingassimpleasaninterestpaymenttoalistofusers.同樣的原因?qū)е逻^(guò)高的I/O操作費(fèi)。例如，一份合約不能對(duì)多個(gè)用戶(hù)使用，因?yàn)檫@一操作的成本超出了區(qū)塊的限制。因此，開(kāi)發(fā)者不得不向用戶(hù)們重復(fù)類(lèi)似于利息支付那樣的簡(jiǎn)單操作。Poordatamodellingtoolsandpoorsupportforqueries.Applicationdevelopershavetoresorttocentralizedindexingandcachinglayers,orusingthirdpartyserviceswhichdonotprovidethesamesecurityguaranteesasthebaselayer.欠佳的數(shù)據(jù)模型工具和查詢(xún)支持。應(yīng)用程序開(kāi)發(fā)者不得不通過(guò)中心索引來(lái)獲取應(yīng)用層，有時(shí)還需借助第三方服務(wù)，但這樣又得不到同等的安全保障。Error--pronecontractlanguagewhichhasresultedinmanyhigh--profile易出錯(cuò)的合約語(yǔ)言已導(dǎo)致許多備受矚目的攔劫Noprovisionforcontractupgradesattheplatformlevel,thisfunctionalityhastobeimplementedasaseparatelayerwhichfurtherincreasescomplexity.在平臺(tái)層面沒(méi)有合約升級(jí)的條款，在單獨(dú)應(yīng)用層實(shí)現(xiàn)這一功能更加劇了復(fù)雜性UsershavingtopayfeesforeveryinteractionresultsinpoorUX.Slowconfirmationsareamajorusabilityissue.用戶(hù)需要為每一次交互付費(fèi)導(dǎo)致了不好的用戶(hù)體驗(yàn)。確認(rèn)速度慢是一個(gè)主要的使用問(wèn)題。Poorlightclientsupport.ThreeyearsafterbeginningdevelopmenteffortstheEthereumFoundationwasstillstrugglingtoofferaproduction--qualitylightwallet.不良的輕客戶(hù)端支持。在開(kāi)發(fā)平臺(tái)三年后，以太坊基金仍舊在研發(fā)產(chǎn)品級(jí)質(zhì)量輕錢(qián)包。Applicationsdesignedwithalargeaudienceinmindneedtobeflexibleandresponsive.Theyrequireaplatformwhichempowersthedevelopertoallocateresourcesinawaythatsuitstheirusers.EvenifEthereumandotherplatformscurrentlyindevelopmenttacklescalabilityissues,theywillnotbeabletoprovideasufficientdegreeofdeveloperautonomyandwillremainasomewhathostileenvironmentfordapps.面向大批量客戶(hù)群的應(yīng)用程序必須具有靈活性和快速響應(yīng)能力。它們要求平臺(tái)賦能開(kāi)發(fā)者完成用戶(hù)所需資源的配置。即使以太坊等其他平臺(tái)目前正發(fā)展解決可擴(kuò)展性的問(wèn)題，他們依舊無(wú)法給開(kāi)發(fā)者提供足夠的自主權(quán)，對(duì)于分散應(yīng)用程序而言仍舊會(huì)是個(gè)糟糕的運(yùn)行環(huán)境。Webelievethattoaddresstheseproblemsproperlyweneedtoseriouslyrethinktheblock--chainarchitectureandprogrammingmodelwiththeneedsofdecentralizedapplicationsin要解決這一系列問(wèn)題，我們需要重新認(rèn)真了解區(qū)塊鏈架構(gòu)，以分散應(yīng)用程序的需求為中心來(lái)設(shè)計(jì)模塊。Blockchainasadatabase作為數(shù)據(jù)庫(kù)的區(qū)塊Themainroleofablockchaininadecentralizedapplicationcontextistomanagedatainasecureandconsistentmanner.Thusitcanbeunderstoodasadatabase,particularly,asasecuredecentralizeddatabase.Anothermajorroleofablockchainispreventionofdouble--spending,butthisisaspecialcaseofdataconsistencyconstraints.在分散應(yīng)用程序的環(huán)境中區(qū)塊鏈的主要職責(zé)是以安全和始終如一的方式管理數(shù)據(jù)。所以它可被定義為一個(gè)數(shù)據(jù)庫(kù)，一個(gè)安全的去中心化數(shù)據(jù)庫(kù)。區(qū)塊鏈的另一個(gè)重要職責(zé)是防止重復(fù)消費(fèi)，但這是針對(duì)數(shù)據(jù)一致性約束的一個(gè)個(gè)例。Blockchainswhichareoptimizedforpayments,suchasBitcoin,canadopthighlyspecialized(andoptimized)datamodels.Butaplatformdesignedforhostingdiversedecentralizedappli--cationsneedsageneral--purposedatamodel.像比特幣這樣優(yōu)化付款的區(qū)塊鏈可以采用高級(jí)定制化（和優(yōu)化）的數(shù)據(jù)模型。但一個(gè)承載多樣化分散應(yīng)用程序的平臺(tái)需要的是一個(gè)多用途的數(shù)據(jù)模型。Mostblockchainplatformsnowadaysusekey--valuedatastores(examples:Ethereum,NEO,Fabric).Thismodelis,intheory,complete,andenablestheuseofhigh--performancedatastoressuchasLevelDB.However,thismodelisverylow--level.Itrequiresapplicationdevel--operstoimplementdetailssuchasserializationandindexing,adauntingchallenge.現(xiàn)行的多數(shù)區(qū)塊鏈平臺(tái)都使用關(guān)鍵值數(shù)據(jù)存儲(chǔ)（如：以太坊，NEO,Fabric等）。從理論上講，這一模型完整且可以使用像LevelDB這樣高效的數(shù)據(jù)存儲(chǔ)。但是，這種模型是非常低層次的。它要求應(yīng)用程序開(kāi)發(fā)者執(zhí)行序列化和檢索等細(xì)節(jié)，這無(wú)疑是一巨大挑戰(zhàn)。Compoundingthis,blockchainplatformstypicallydonotexposethefullfunctionalityofkey--valuestores,suchastheabilitytousearbitrary--sizedkeysanditeratethroughkeys.Forex--ample,inEVMallkeysare256--bitintegersanditeratingthroughstoredkeysisimpossible.Forthesereasons,implementingproperindexeddataaccessontheEVMisbothdifficultand綜合以上情況，區(qū)塊鏈平臺(tái)無(wú)法展示關(guān)鍵值存儲(chǔ)的所有功能，比如任意大小值和重復(fù)值的使用能力。在EVM中所有值都是256比特整，重復(fù)使用存儲(chǔ)值也是無(wú)法實(shí)現(xiàn)的。為此，在上執(zhí)行適當(dāng)?shù)臋z索數(shù)據(jù)權(quán)限是困難且無(wú)效的。Relationalmodel關(guān)系模Forthereasonsstatedabove,weconsiderourchoiceofdatamodeltobethelynchpinofourblockchainplatform.綜上所述原因，我們將數(shù)據(jù)模型的選擇視為區(qū)塊鏈平臺(tái)的關(guān)鍵Therelationalmodelhasbeenthegoldstandardfordatabasemanagementforthelastfivedecades.Rootedinmathematicsandlogic,itisknowntobeabletomodelcomplexdatainanefficientway.關(guān)系模型在過(guò)去的五十年里都是數(shù)據(jù)庫(kù)管理的黃金標(biāo)準(zhǔn)。從數(shù)學(xué)和邏輯角度出發(fā)，關(guān)系模型被認(rèn)為可以高效建立復(fù)雜數(shù)據(jù)模型。Asdecentralizedapplicationsdealwithincreasinglycomplexdatastructures,thepoweroftherelationalmodelbecomesmoreandmoreapparent.Further,mostsoftwareengineersarealreadyfamiliarwithitsotheywon’thavetolearnnewconceptsinordertoimplementan由于分散應(yīng)用程序處理著日益復(fù)雜的數(shù)據(jù)架構(gòu)，關(guān)系模型的力量也日趨明顯。此外，大多數(shù)軟件工程師都已熟悉關(guān)系模型，無(wú)需再為執(zhí)行應(yīng)用程序?qū)W習(xí)新的理念。ArelationalmodelalsoallowsustoleveragethepowerofSQLdatabasemanagementsys--temswhichhavebeenoptimizedfordecades.Insteadofdappcodewhichtraversesmemorycellsonebyone,wecansendaquerytotheDBMSandletituseitssophisticatedqueryplanning,datastructuresandcachingcapabilitiestocarryoutthequeryasfastaspossible.關(guān)系模型也使我們能利用已經(jīng)被優(yōu)化了幾十年的SQL數(shù)據(jù)庫(kù)管理系統(tǒng)。與其讓分散應(yīng)用程序代碼一個(gè)個(gè)穿過(guò)記憶細(xì)胞，我們可以向DBMS發(fā)送一個(gè)查詢(xún)請(qǐng)求，讓它運(yùn)用其精細(xì)的查詢(xún)?cè)O(shè)計(jì)、數(shù)據(jù)架構(gòu)和獲取能力來(lái)快速開(kāi)展查詢(xún)。Ofcourse,thechoiceofdatamodelisatradeoff.Therelationalmodelmighthavethefollow--ingdisadvantages:當(dāng)然，數(shù)據(jù)模型的選擇也是一個(gè)折中的考量。關(guān)系模型也有以下的一些缺陷Performanceishardtopredictanddependsonthequeryplanner.Thisisnotasignif--icantdisadvantageinthecontextofChromiabecauseeachdappwillberuninanisolatedmanner;?slowquerieswillaffectonlythedappwhichperformsthemratherthanthesystemasawhole.性能的好壞很難預(yù)測(cè)，它取決于查詢(xún)計(jì)劃者。這對(duì)于Chromia也并非是一個(gè)重大缺陷，因?yàn)槊恳粋€(gè)分散應(yīng)用程序都以單獨(dú)的方式運(yùn)行，查詢(xún)緩慢只會(huì)影響分散應(yīng)用程序，而不是系統(tǒng)本身。Itisimpossibletoimposehardboundsonqueryexecutiontime.Again,thisisnotaprobleminthecaseofChromiabecauseitaffectsonlytheapplicationwhichissuesslowqueries.對(duì)查詢(xún)執(zhí)行時(shí)間規(guī)定明確界限并不切實(shí)可行。同樣，這對(duì)Chromia而言也不是問(wèn)題，ParallelizationofSQLdatabasesisacomplexareaofactiveresearch.Asfarasweknow,noblockchainplatformoffers100%fullyautomaticparallelizationonamassivescale.Thusthereisnoevidencethatarelationalmodelisworsethanothermodels.Inaddition,webelievethattherelationalmodelwillmakelogicalshardingandsidechainmechanismseasiertoimplement.SQL數(shù)據(jù)庫(kù)的平行化是有效調(diào)查的一個(gè)難點(diǎn)。據(jù)目前了解，尚未有區(qū)塊鏈平臺(tái)能在大范圍內(nèi)提供100%自動(dòng)平行化。所以，沒(méi)有證據(jù)可以證明關(guān)系模型不如其他模型。另外，我們認(rèn)為關(guān)系模塊將會(huì)使邏輯分片和側(cè)鏈機(jī)制更易執(zhí)行。First--classdecentralizedapplications頂尖的分散應(yīng)用程InEthereumallcodelivesin“contracts”.Itdoesnotdifferentiatebetweenindividualwalletcontractsandcomplexmulti--usercontracts,theyallusethesameresourcemeteringnadpro--grammingmodel.AnEthereum--baseddappwilluseoneormorecontracts(possiblyacontractforeachuser)andfront--endcomponents.Infact,manyEthereumapplicationsmakeuseofcentralizedcaching,renderingtheir“decentralized”credentialssomewhatdubious.在以太坊上，所有的代碼存活于“合約”中。私有錢(qián)包合約與復(fù)雜多用戶(hù)合約并未區(qū)分開(kāi)，他利用的是同一資源測(cè)量程序模型。一個(gè)基于以太坊的分散應(yīng)用程序會(huì)用一個(gè)或多個(gè)合約（一個(gè)用戶(hù)一份合約）和前端組件。事實(shí)上，許多以太坊應(yīng)用程序利用中心化緩存來(lái)補(bǔ)償“去中心化”證書(shū)，這多少有些可疑。Whilethisapproachisquiteelegantandcanscaletodifferentkindsofapplications,itisveryinconvenientfordappsdesignedformassuse.End--usershavetopayforeveryinteractionwiththeirdapp,inproportiontothecomputationalandstorageresourcesrequiredfortheirtransaction.Inotherwords,Ethereumdoesn’tgivedecentralizedapplicationstheflexibilitytomanageresourcesthemselves.Forexample,a“freemium”businessmodelisoutrightimpos--sible.Thiscreatesabarrierfordecentralizedapplicationadoption:mostusersarenotreadytopayforeverysingleclick.雖然這一方法十分明智，可以分布到各不同種類(lèi)的應(yīng)用程序，但對(duì)于大范圍使用的分散應(yīng)用序來(lái)說(shuō)很不方便。終端用戶(hù)需要為他們分散應(yīng)用程序上的每一次交互付費(fèi)，與他們所需的計(jì)算和存儲(chǔ)資源成正比。換言之，以太坊沒(méi)有給分散應(yīng)用程序該有的靈活性來(lái)管理他們自己的資源。例如，“免費(fèi)增值”的商務(wù)模型是完全無(wú)法實(shí)現(xiàn)的。這給分散應(yīng)用程序的使用添加了一道屏障，大多數(shù)用戶(hù)并未準(zhǔn)備好為他們的每一次操作買(mǎi)單。ChromiasolvesthisissuebyprovisioningresourcesonthedecentralizedapplicationChromia通過(guò)在分散應(yīng)用程序?qū)用嫣峁┵Y源來(lái)解決此問(wèn)題Eachdapphasitsownblockchain每個(gè)分散應(yīng)用程序擁有自己的區(qū)塊鏈（側(cè)鏈Fees(collectedtomaintainnodes)arepaidbythedappasawhole,notbyend--users由分散應(yīng)用程序統(tǒng)一付費(fèi)，而非終端用Thusdappsarefreetoimplementtheirownresourcemanagementpolicies,whichcanthusbealignedwitheconomicratherthantechnicalneeds分散應(yīng)用程序能夠執(zhí)行自己的資源管理政策，可以更好的統(tǒng)一經(jīng)濟(jì)需求，而非技術(shù)需求。Everyblockchainneedsananti--spammechanism,butthismechanismdoesn’thavetobetiedtofees.Forexample,adappmightallowonly1actionfromausereach15seconds,thusasingleuserwon’tbeabletospamtheblockchainwithbillionsoftransactions.AdappcanalsomitigateSybilattacksthroughlimitingnewuserregistrationtosomereasonablerateand/orrequiringinvitationoradeposit.每個(gè)區(qū)塊鏈都需要一個(gè)反垃圾信息機(jī)制，但這一機(jī)制并非要與費(fèi)用掛鉤。例如，一個(gè)分散應(yīng)程序可能在15秒內(nèi)只允許用戶(hù)有一次操作，因此單一用戶(hù)不會(huì)在區(qū)塊鏈上有數(shù)十億次交易。分散應(yīng)用程序還可通過(guò)一定程度限制新用戶(hù)注冊(cè)量或通過(guò)邀請(qǐng)、押金的方式注冊(cè)來(lái)減輕Sybil攻擊。Inthismodel,wedonotneedtomeasuretheresourcesusedbyeachoperation.Instead,weprovisionresourcestotheapplicationasawhole:eachdapp’sblockchainwillrunonacertainsetofnodes.TypicallyitwillhaveitsowndedicatedCPUthread.在這一模型下，我們無(wú)需衡量每一操作的資源。相反，我們將資源整體運(yùn)用到應(yīng)用程序上：每一分散應(yīng)用程序的區(qū)塊鏈通過(guò)一定數(shù)量的節(jié)點(diǎn)運(yùn)行。它將有自己專(zhuān)屬的CPU螺紋。Thisremovesresourcemeteringoverhead(wenolongercarehowmanyinstructionswereexecuted,asanapplicationcannotusemoreresourcesthanitwasgiven)allowingdappstoperformfasterandscalebetter.這一資源轉(zhuǎn)移使得分散應(yīng)用程序運(yùn)行更快，延伸更廣（我們不用再關(guān)注執(zhí)行了多少指令，因?yàn)閼?yīng)用程序已最大化使用其資源）。Ifadappneedsmorethanoneexecutionthread,itcanconsistofmultipleshardseachofwhichwillbeasidechain.如果一個(gè)分散應(yīng)用程序需要多個(gè)執(zhí)行螺紋，它可包含多個(gè)分片，即側(cè)鏈Besidesscheduling,havingdappsasfirst--classcitizenontheplatformallowsthefollowing除了調(diào)度，作為平臺(tái)上的頂尖市民，分散應(yīng)用程序擁有以下特征Tokeneconomicsintegratedwithafeemodel,i.e.feesaretakenfromprofits“earned”byanapplication代幣經(jīng)濟(jì)與費(fèi)用模型融合，如程序的支出可在其所獲盈利中支取Built--ingovernance內(nèi)置管Updates更Programmingmodel編程模ThePostchainframeworkonwhichChromiaisbasedallowsustouseexistingopensourceSQLdatabasesoftware(specifically,PostgreSQL)toimplementdatastoreandquerycapa--bilities.However,wecannotallowdappstoperformarbitrarySQLqueriesassaidqueriesmightbeunsafe,ambiguousorleadtoexcessiveresourceuse.Chromia所在的后鏈框架允許我們利用現(xiàn)有開(kāi)放的SQL數(shù)據(jù)庫(kù)軟件來(lái)執(zhí)行數(shù)據(jù)存儲(chǔ)和查詢(xún)。但是，我們不允許分散應(yīng)用程序隨意執(zhí)行SQL查詢(xún)，因?yàn)椴樵?xún)可能不安全、模糊不清或?qū)е沦Y源濫用。Mostdappblockchainplatformsusevirtualmachinesofvariouskinds.Butatraditionalvirtualmachinearchitecturedoesn’tworkverywellwiththeChromiarelationaldatamodel,asweneedawaytoencodequeriesratherthanjustoperations.大多數(shù)分散應(yīng)用程序區(qū)塊鏈平臺(tái)使用各種虛擬機(jī)器。但是傳統(tǒng)的虛擬機(jī)器在Chromia關(guān)系數(shù)Forthisreason,wearetakingmorelanguage--centricapproach:anewlanguagecalledRell([Rel]ational[l]anguage)willbeusedfordappprogramming.Thislanguageallowsprogram--merstodescribe:為此，我們采取更多以語(yǔ)言為中心的方法：一種叫做Rell（關(guān)系語(yǔ)言）的新語(yǔ)言將應(yīng)用于分散應(yīng)用程序的編程。這一語(yǔ)言使程序員能描述：Schema/datamodel模式/數(shù)據(jù)模Queries查Proceduralapplicationcode程序代Rellwillbecompiledtoanintermediatebinaryformatwhichcanbeunderstoodascodeforaspecializedvirtualmachine.ChromianodeswillthentranslatequeriescontainedinthiscodeintoSQL(whilemakingsurethistranslationissafe)andexecutecodeasneededusinganinterpreterorcompiler.Rell會(huì)編制一個(gè)二進(jìn)制格式，可理解為一個(gè)專(zhuān)屬虛擬機(jī)器的代碼。Chromia的節(jié)點(diǎn)就會(huì)將帶此代碼的查詢(xún)翻譯成SQL，用編譯器來(lái)執(zhí)行代碼RellwillhavethefollowingRell有以下特征Typesafety/statictypechecks.It’sveryimportanttocatchprogrammingerrorsatthecompilationstagetopreventfinanciallosses.Rellwillbemuchmoretype--safethanSQL,anditwillmakesurethattypesreturnedbyqueriesmatchtypesusedinproce--duralcode.輸入安全/靜態(tài)類(lèi)型檢查。在編輯階段發(fā)現(xiàn)編程錯(cuò)誤很重要，可以防止不必要的經(jīng)濟(jì)損失。Rell比SQL在編寫(xiě)時(shí)更加安全，它可確保查詢(xún)反饋輸入與代碼輸入相匹配。Safety--optimized.Arithmeticoperationsaresaferightoutofthebox,programmersdonotneedtoworryaboutoverflows.Authorizationchecksareexplicitlyrequired.安全優(yōu)化。算術(shù)操作安全可用，程序員無(wú)需擔(dān)心超限。授權(quán)檢查是明確要求的Concise,expressiveandconvenient.ManydevelopershateSQLbecauseit’sveryverbose.Relldoesn’tbotherdeveloperswithdetailswhichcanbederivedautomati--cally.Asadatadefinitionlanguage,Rellisupto7xmorecompactthanSQL.精準(zhǔn)、可表現(xiàn)、方便。許多開(kāi)發(fā)者不喜歡SQL，因?yàn)樗捌浞爆?。Rell不會(huì)過(guò)多打擾開(kāi)發(fā)者，細(xì)節(jié)之處可自動(dòng)獲取。作為一個(gè)數(shù)據(jù)定義語(yǔ)言，Rell比SQL簡(jiǎn)潔7倍多。Allowsmeta--programming.Wedonotwantapplicationdeveloperstoimplementthebasicsfromscratchforeverydapp.Rellwillallowfunctionalitytobebundledastem--支持元編程。我們不想讓?xiě)?yīng)用程序開(kāi)發(fā)者忙于為每一個(gè)分散應(yīng)用程序執(zhí)行最基礎(chǔ)的內(nèi)容。所以，Rell支持模板功能。Weidentifiedthatnoexistinglanguageorenvironmenthasthefeaturesetrequiredforthistask,andthusdevelopmentofanewlanguageisabsolutelynecessary.我們意識(shí)到現(xiàn)存的任何語(yǔ)言或環(huán)境都無(wú)法完成此任務(wù)，所以開(kāi)發(fā)一種新的語(yǔ)言迫在眉睫WedesignedRellinsuchawaythatitiseasytolearnfor我們?cè)O(shè)計(jì)的Rell易于程序員學(xué)習(xí)Programmerscanuserelationalprogrammingidiomstheyarealreadyfamiliarwith.However,theydon’thavetogooutofwaytoexpresseverythingthroughrelationalalgebra:Rellcanseamlesslymergerelationalconstructswithproceduralprogram--程序員可以運(yùn)用他們已經(jīng)熟悉的關(guān)系編程語(yǔ)言。但是他們無(wú)需用關(guān)系代數(shù)來(lái)表述所有東西。Rell可以將關(guān)系架構(gòu)與程序編程完美融合。ThelanguageisdeliberatelysimilartomodernprogramminglanguageslikeJavaScriptandKotlin.Afamiliarlanguageiseasiertoadaptto,andourinternaltestsshowthatprogrammerscanbecomeproficientinRellinmatterofdays.Incontrast,theALGOL--stylesyntaxofPL/SQLgenerallyfeelsancientandweirdtomoderndevelopers.該語(yǔ)言和JavaScript和Kotlin這樣的編程語(yǔ)言非常相似。一種熟悉的語(yǔ)言更容易被采納，我們的內(nèi)部測(cè)試顯示程序員幾天內(nèi)就可熟練運(yùn)用Rell。相反，PL/SQL使用的ALGOL型語(yǔ)法顯得和當(dāng)今行業(yè)有些格格不入。TheEthereumprogrammingmodelistypicallydescribedasveryerror--prone.BugsinEthereumsmartcontractshaveresultedinlossestotallinghundredsofmillionsofdollars7.InChromia,weaimtoeliminatemostcommonsourcesofproblemsthroughabetterprogram--mingmodel(noweirdinteractionsbetweendifferentsmartcontractsasintheDAOcase89)andsaferlanguages.以太坊編程模型被認(rèn)為極易出錯(cuò)。以太坊智能合約中的故障已導(dǎo)致總計(jì)數(shù)億美元的損失。Chromia，我們希望通過(guò)更優(yōu)的編程模型和更安全的語(yǔ)言來(lái)消除這些最常見(jiàn)問(wèn)題的根源7AlistofthemostseriousEthereumvulnerabilitiescanbefoundhere可訪(fǎng)問(wèn)以下鏈接找到最嚴(yán)重的以太坊脆弱性清單:https://www.dasp.co/8hpsbogehereumorg20160617crcal-update--re--dao--9hpswwwwredcom20160650-million--hack--just--showed--dao--OnEthereumcodeisimmutable,itisoftenimpossibleforadevelopertofixherdappunlesssheretainsfullcontrol,thusmakingitnot--quite--decentralized.InChromia,upgradescanbedeployedthroughabuilt--ingovernanceandtransitionmechanism.在以太坊上的代碼是不可變的，開(kāi)發(fā)者想要處理分散應(yīng)用程序必須保持對(duì)程序的完全控制，這樣就做不到去中心化了。在Chromia，程序更新可以通過(guò)內(nèi)置的管理和轉(zhuǎn)變機(jī)制來(lái)完成。Consensus&nodes共識(shí)&節(jié)ModeloverviewItisclearthatthefullnodemodeldoesn’tscaleparticularlywell.Ifwerequireuserstorunafullnodewhichhasacompletecopyofthesystemstatethendappsareseverelylimitedinwhatcomputationsandstorageresourcestheycanuse.很顯然，全節(jié)點(diǎn)模型并沒(méi)有延伸的很好。如果我們要求用戶(hù)運(yùn)行擁有系統(tǒng)狀態(tài)全部備份的全節(jié)點(diǎn)，那分散應(yīng)用程序可用的計(jì)算和存儲(chǔ)資源就變得極其有限。Withtheaimofachievingbetterperformanceatscaleweproposeamodelinwhichindividualdappsarehostedonasubsetofvalidatornodes,whichestablishconsensusonanymodifi--cationstothedappstate,andhandleclientqueries.Thesystemshouldpermitanyusertorunafullreplicanodeifdesired,butthesystemshouldnotdependonthesereplicanodesfor為了能在延伸性方面能有更佳表現(xiàn)，我們推出了讓單一分散應(yīng)用程序在驗(yàn)證器子節(jié)點(diǎn)上運(yùn)行的模型，這樣對(duì)分散應(yīng)用程序狀態(tài)的任何修改都可達(dá)成共識(shí)，也可處理客戶(hù)查詢(xún)。系統(tǒng)應(yīng)該允許任何用戶(hù)運(yùn)行想要的全復(fù)制節(jié)點(diǎn)，但不應(yīng)該將常規(guī)運(yùn)行依賴(lài)于此。SybilcontrolmechanismSybilTheresearchdonebyourteamindicatesthatcommonlyusedSybilcontrolmechanismslikePoWandProofofStake(PoS)areunsatisfactory10:neitherofthemguaranteesasufficientlevelofSybilattackmitigation,orevenaparticularlygoodmeasureofdecentralization.Evi--denceindicatesthatmostPoW--basedblockchains,includingBitcoin,mightbedefactocon--trolledbyasmallgroupofentities.Thisproblemisparticularlybadforsmallercryptocurrencieswhichdonotyethaveanindependentminingecosystem.PoSalsocomeswithnodecentral--izationguarantees,andDPoS11inparticularispronetoformationofcartelsandbribery.調(diào)查發(fā)現(xiàn)，像PoW和PoS這樣常見(jiàn)的Sybil控制機(jī)制表現(xiàn)都不盡如人意。它們都無(wú)法保證Sybil攻擊有一定層級(jí)的減輕，甚至好的去中心化方式都沒(méi)有。事實(shí)證明，包括比特幣在內(nèi)的多數(shù)基于PoW的區(qū)塊鏈都很可能是由一小群個(gè)體控制的。這一問(wèn)題對(duì)于小型加密貨幣尤為突出，因?yàn)樗鼈儧](méi)有自己獨(dú)立的挖礦生態(tài)系統(tǒng)。PoS也沒(méi)有去中心化保障，DPoS甚至有壟斷和賄賂傾向。ThusinsteadoffollowingcommonlyusedapproacheswewilldesignChromiaconsensusandSybilcontrolmechanismsfromfirstprinciples.1011DelegatedProofofStake股份授權(quán)證明機(jī)制,hpsbsharesorgechnoogydeegaed-proof--of--因此，摒棄過(guò)往常用的方法，我們?cè)O(shè)計(jì)出了Chromia共識(shí)和Sybil控制機(jī)制FirstweobservethatwhatChromiaistryingtoachievecanbecomparedtocloudcomputing:anapplicationwhichredundantlyusesmultiplecloudhostingproviderscanbeconsideredadecentralizedapplication,inthesensethatfailureorcensorshipofasinglecloudhostingpro--viderdoesnotresultinashutdownofthewholeapplication.Acloudcomputingmodelalsoallowsuserstousethinclientsinsteadofhostingacompletereplicaofapplicationbackendontheirpersonaldevice.首先，我們發(fā)現(xiàn)Chromia所想要達(dá)成的可比作為云計(jì)算：一個(gè)過(guò)多使用多個(gè)云代管供應(yīng)商的應(yīng)用程序可被認(rèn)為是一個(gè)分散應(yīng)用程序。單一云代管供應(yīng)商的故障或?qū)彶椴粫?huì)導(dǎo)致整個(gè)應(yīng)用程序的癱瘓。云計(jì)算模型也使得用戶(hù)可以使用瘦客戶(hù)，而不是在他們個(gè)人設(shè)備上運(yùn)行一整個(gè)應(yīng)用程序后端。TheessentialrolesintheChromiamodelaredefinedasfollows.Chromiasoftwarerunsonnodes,physicalorvirtualinstancesofcomputingpower.Nodesarecontrolledorperhapsownedbysomekindofindividual,organisation,orcollectivewhichwerefertoasaprovider.Usersconnecttosuchnodestoposttransactions,querydataorsynchronizetheirprivateChromia模型定義的基本職責(zé)如下。Chromia軟件在節(jié)點(diǎn)上運(yùn)行，這些節(jié)點(diǎn)是物理或虛擬的運(yùn)算力。節(jié)點(diǎn)被一些個(gè)人、組織或集體所控制或擁有，我們將他們稱(chēng)之為供應(yīng)商。用戶(hù)連接這些節(jié)點(diǎn)以完成盤(pán)后交易、數(shù)據(jù)查詢(xún)或同步私人備份。AByzantinefaulttolerantnetworkisdistinguishedfromamerelyfaulttolerantnetworkbyitsabilitytotoleratearbitraryandpotentiallymaliciousbehaviourbynetworkparticipants.Theconceptofnodesissufficientfordesigningafaulttolerantnetwork,buttotargetproperByz--antinefaulttolerancewemust