Where Blockchains Add Real Value

WHERE BLOCKCHAINS ADD REAL VALUE

GIDEON GREENSPAN

Many different uses have been proposed for blockchains: tracking financial assets, sharing health records, managing identities, coordinating supply chains. However, many such proposals have suffered from a lack of critical and system- atic thinking, thus they fail to either take advantage of the key benefits of blockchains or take note of their disadvantages.

At the simplest level, a blockchain is a on the raison d’être of blockchains, data- new way of constructing a database in base disintermediation, which can be which the database is placed under the defined as the ability for multiple parties collective control of multiple parties to directly share a single database without instead of a central authority. Therefore, putting that database under a single the first question to ask when assessing a party’s control. We provide a checklist for way of using blockchain is whether it assessing whether this disintermediation could be satisfactorily implemented on a is helpful. We next look at two key disad- regular centralized database, such as vantages of blockchains when compared Oracle, SQL Server, MySQL, or Postgres. with regular databases—performance and If the answer is “yes,” there is no value in confidentiality. We then outline four gen- using a blockchain, which remains a rela- eral types of usage in which the tradeoffs tively young and immature technology, tend to favor a blockchain architecture. whereas programs like Oracle and Finally, we look at three real-world exam- MySQL have decades of development ples of our software being used in produc- behind them. tion to see what conclusions can be Nonetheless, there are certainly a drawn. good number of applications whose ideal It’s important to clarify that this architecture has a blockchain at its core. paper is written primarily with “permis- This paper provides a framework for eval- sioned” blockchains in mind, which are uating the viability of blockchain usage fundamentally different from the permis- from four different angles. First we focus sion-less blockchains that underlie cryp-

58 innovations / Blockchain for Global Development

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 tocurrencies such as Bitcoin and cious behavior of a minority of these val- Ethereum, although they share many idators, they can be undermined by a val- technical characteristics. Permissionless idator majority, whose members would blockchains like Bitcoin need economic face no immediate economic cost. consensus mechanisms such as proof-of- Nonetheless, a permissioned validation work or proof-of-stake to regulate the scheme has the advantage of not requir- process, and anyone who makes the nec- ing complex economic incentives and essary investment can participate. These crypto tokens, and thus is suitable in situ- mechanisms provide the key benefit of ations where the validators are genuinely making it financially costly to attempt to motivated by a simple and common undermine the chain’s functioning. interest in maintaining a functioning sys- However, the corollary is the need to use tem. Moreover, permissioned an associated crypto token, which may blockchains are often double permis- draw suspicion from governments and sioned, which means that only certain financial institutions, and whose value designated parties are able to connect and tends to be highly volatile. transact on them.1. Unlike cryptocurrencies, the consen- sus mechanism used in permissioned DO YOU NEED A blockchains relies on a federation of iden- BLOCKCHAIN? tified validators who form a majority con- sensus about the blockchain’s transac- Let us next examine the most basic reason tions. While permissioned blockchains for using a blockchain in a project—the are resistant to an error by or the mali- need for database disintermediation. The

ABOUT THE AUTHOR Gideon Greenspan is Founder and CEO of Coin Sciences Ltd., the company that developed the MultiChain, the platform for permissioned blockchains. MultiChain includes features such as permissions management, native assets, data streams and simple configuration and deployment, and it has been used successfully for blockchain projects in many of the world’s largest banks, consulting firms, financial technology, and IT companies. Prior to founding Coin Sciences, Greenspan conceived and built many profitable web servic- es with a deep algorithmic element, including Copyscape, the plagiarism search engine, and Web Sudoku, a popular sudoku website. Greenspan holds a PhD in computer science from Technion, an MA in philosophy from King’s College London, and an MA in computer science and management studies from Cambridge University. This paper is based on an edited collection of four posts published on the MultiChain blog: 1. https://www.multichain.com/blog/2015/11/avoiding-pointless-blockchain-project/ 2. https://www.multichain.com/blog/2016/03/blockchains-vs-centralized-databases/ 3. https://www.multichain.com/blog/2016/05/four-genuine-blockchain-use-cases/ 4. https://www.multichain.com/blog/2017/11/three-non-pointless-blockchains-production/ © 2018 Gideon Greenspan

innovations / volume 12, number 1/2 59

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Gideon Greenspan

following five aspects of blockchains can peer-to-peer fashion. However, transac- help you determine whether this need tions might also be created by users who exists. are not running a node. Consider, for example, a payment system that is main- The Database tained collectively by a small group of banks but has millions of end users on Blockchains are a technology to be used mobile devices who communicate only with shared databases. The first step in with their own bank’s systems. deciding whether to use one is to know why you are using a database, which can A Lack of Trust be defined as a structured repository of information. This repository can be a tra- Blockchains are a technology for databas- ditional relational database, which con- es that have multiple writers, and where tains one or more spreadsheet-like tables, there is some degree of mistrust among or it can be of the trendier NoSQL variety, the multiple entities writing to a given which works more like a file system or database. dictionary. It might appear that mistrust arises For example, a ledger for financial only between organizations, such as assets can be expressed naturally as a banks trading in a marketplace or compa- database table in which each row repre- nies involved in a supply chain. However, sents one type of asset owned by one par- it also can exist between departments ticular entity. Each row has three columns within a single organization, for example, that contain (1) the owner’s identifier, or between the organization’s operations such as an account number; (2) an identi- in different countries. fier for the asset type, such as USD or Mistrust in a database context means AAPL; and (3) the quantity of that asset that one user is not willing to let another held by that owner. modify the database entries they “own.” Databases are modified via transac- Similarly, one user will not accept as tions that involve a set of changes to the gospel the “truth” another user reports in database, which must be accepted or their database because each has different rejected as a whole. In the case of an asset economic or political incentives. ledger, a payment from one user to anoth- er involves a transaction that deducts the Disintermediation appropriate quantity from one row and adds it to another. The problem, as defined so far, is how to operate a database that has multiple mis- Multiple Writers trusting writers. This problem already has a well-known solution: the trusted inter- For a blockchain to be worth considering, mediary, who can be a party all the writers the database in question must have multi- trust, even if they don’t fully trust each ple writers—that is, more than one entity other. The world is filled with databases must be generating the transactions that of this nature, such as the ledger of modify the database. It must be possible accounts in a bank. The bank controls the to draw up a list of the identities of these database and ensures that every transac- writers. tion is validated and authorized by the In most cases, the writers will also customer whose funds it moves. No mat- operate “nodes” that hold a copy of the ter how politely the customer asks, the database, process transactions locally, and bank will never let them modify the data- relay the transactions to other nodes in a base directly.

60 innovations / Blockchain for Global Development

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Where Blockchains Add Real Value

Blockchains remove the need for is that transactions can be created collab- trusted intermediaries by enabling the oratively by multiple writers without direct modification of databases with either party exposing themselves to risk. multiple mistrusting writers. No central This is what allows the safe “delivery ver- gatekeeper is required to verify transac- sus payment” settlement of financial tions and authenticate their source. transactions using a blockchain without a Instead, the definition of a transaction is trusted intermediary.2. extended to include proof of authorization Another good case for using and proof of validity. Transactions there- blockchains can be made for situations in fore can be independently verified and which transactions from different writers processed by every node that maintains a are independent but cross-correlated with copy of the database. each other at the time of reading by any Logical questions to ask are, Does node. One example is a shared-identity my application need this disintermedia- database in which multiple entities vali- tion? Is there anything wrong with having date different aspects of consumers’ iden- a central party maintain an authoritative tities. Although each such certification database and act as the transaction gate- stands alone, the blockchain provides a keeper? The answer is, there may be good useful way to bring everything together in reasons to prefer a blockchain-based data- a unified way. base over a trusted intermediary. These include lower costs, a faster workflow, BLOCKCHAIN DRAWBACKS automatic reconciliation, new regula- tions, or the inability to find a suitable Let’s assume that, having reviewed the intermediary. checklist above, you have a genuine prospect for using a blockchain because Transaction Interaction database disintermediation will provide a key business benefit. However, before you As explained so far, blockchains make conclude that the project should be built sense for databases that are shared by on the blockchain, it’s important to multiple writers who modify the database understand the two key disadvantages of directly but don’t entirely trust each doing so—loss of confidentiality and other. Blockchains become even more reduced performance. powerful when there is some interaction between the transactions the multiple Confidentiality writers create. Transaction interactions mean that Every node in a blockchain independently the operations performed by different verifies and processes every transaction, writers on the database often depend on without relying on the opinion of the one other. Let’s say, for example, that other nodes. A node can do this because it Alice sends some funds to Bob, and Bob has full visibility into (a) the database’s then sends some on to Charlie. In this current state, (b) the modification case, Bob’s transaction is dependent on requested by a transaction, (c) the rules Alice’s, and there’s no way to verify governing legitimate transactions, and (d) Bob’s transaction without checking a digital signature that proves each trans- Alice’s first. Because of this dependency, action’s origin. While this is undoubtedly the transactions naturally belong in a sin- a clever new way to construct a database, gle shared database. it has a key downside: for many applica- An additional benefit of blockchains tions, especially financial, the full trans-

innovations / volume 12, number 1/2 61

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Gideon Greenspan

parency enjoyed by every node is an ular database does, but it carries three absolute deal killer. additional burdens: Systems built on a regular centralized 1. Signature verification. Every database do not have this problem. blockchain transaction must be digitally Although they too restrict the transac- signed using a public-private cryptogra- tions particular users can perform, these phy scheme such as ECDSA. This is nec- restrictions are imposed in one central essary because transactions propagate location. As a result, the full database con- between nodes in a peer-to-peer fashion, tents need be visible only at that location, so their source cannot otherwise be rather than in multiple nodes. Requests to proven. The generation and verification of read data also go through this central these signatures are computationally authority, which can accept or reject them complex and they create the primary bot- as it sees fit. Whereas a regular database is tleneck in many blockchain platforms. By both read-controlled and write-con- contrast, once a connection has been trolled, a blockchain can be write-con- established in a centralized database, trolled. there is no need to individually verify To be fair, many strategies are avail- every request that comes over it. able to mitigate this problem. Some are 2. Consensus mechanisms. In a dis- simple ideas, such as transacting under tributed database such as a blockchain, an multiple blockchain identities or encrypt- effort must be made to ensure that nodes ing information that must only be visible in the network reach consensus. to certain parties. For financial use, where Depending on the consensus mechanism nodes have to validate payments without used, this might involve significant back- knowing all of the details, advanced cryp- and-forth communication and/or dealing tographic techniques such as confidential with temporary breaks in consensus transactions and zero-knowledge proofs (“forks”) and their subsequent resolution. are also being developed.3.,4. We shall not While centralized databases also must go into the details of these techniques contend with conflicting and aborted here. In a general sense, the more infor- transactions, they are far less likely to mation you want to hide on a blockchain, occur where transactions are queued and the heavier the computational burden you processed in a single location. will pay to generate and verify transac- 3. Redundancy. Putting aside the tions. And no matter how these tech- performance of an individual node, one niques develop, they will never beat the should also consider the total amount of simple and straightforward method of computation a blockchain requires. completely hiding data within a single Whereas centralized databases process trusted intermediary. transactions once (twice in a single repli- cated backup system), blockchain trans- Performance actions must be processed independently by every node in the network. Therefore, A second disadvantage of blockchains is keeping the database updated requires that they will always be slower than cen- much more work. tralized databases. It’s not just that today’s blockchains are slow because the technol- USAGE TEMPLATES ogy is new and unoptimized, it’s due to the nature of blockchains themselves. Let us put questions of performance When processing transactions, a block- aside, as they can be solved for most chain has to do all the same things a reg- blockchain uses by deploying sufficient

62 innovations / Blockchain for Global Development

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Where Blockchains Add Real Value

resources. From our experience in assess- Lightweight Financial Systems ing blockchain applications, the central Let’s start with the class of blockchain tradeoff between blockchains and central- application that will be most familiar— ized databases can be summarized as fol- that in which a group of entities wants to lows: set up a financial system. Within this sys- 
Disintermediation. Blockchains enable tem, one or more scarce assets will be multiple parties who do not fully trust transacted and exchanged between those each other to share a single database entities. safely and directly without requiring a In order for any asset to remain trusted intermediary. 
Confidentiality. All participants in a scarce, two related problems must be solved. First, it must be ensured that the blockchain see all of the transactions same unit of the asset cannot be sent to taking place. Even if various techniques more than one place (a “double spend”). are used to hide some aspects of a trans- Second, it must be impossible for anyone action, a blockchain will always leak to create new units of the asset on a whim more information than a centralized (“forgery”). Any entity able to do either of database. these things could steal unlimited value In sum, blockchains are ideal for from the system. shared databases in which every user is A common solution to these prob- able to read everything but no single user lems is physical tokens, such as metal controls who can write what. By contrast, coins or those securely printed on paper. with traditional databases, a single entity These tokens solve the problem of double exerts control over all read and write spending because the rules of physics (lit- operations, and other users are entirely erally) prevent one token from being in subject to that entity’s whims. two places at the same time. The problem When do these tradeoffs favor using a of forgery is solved by making the tokens blockchain? This question can be extremely difficult to manufacture. Still, approached both theoretically and empir- physical tokens suffer from several short- ically; theoretically by focusing on the key comings that can render them impracti- differences between blockchains and tra- cal: ditional databases, and on how these dif- 
As pure bearer assets, physical tokens ferences inform the possible uses; and empirically, at least in our case, by catego- can be stolen with no way to trace them rizing the real-world solutions being built and without recourse. on the MultiChain platform. Not surpris- 
Physical tokens are slow and costly to ingly, whether we focus on theory or move in large amounts and/or over practice, the same usage categories arise: long distances. 
Lightweight financial systems 
It is tricky and expensive to create phys- 
Provenance tracking ical tokens that cannot be forged. 
Interorganizational record-keeping These shortcomings can be avoided by leaving physical tokens behind and 
Multiparty aggregation redefining asset ownership in terms of a Below we examine the four types of ledger managed by a trusted intermedi- usage in the light of the core tradeoff. We ary. In the past these ledgers were based explain for each why the benefit of disin- on paper records, and today they tend to termediation outweighs the reduced con- run on regular databases. Either way, the fidentiality. intermediary transfers ownership by modifying the ledger’s content in

innovations / volume 12, number 1/2 63

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Gideon Greenspan

response to an authenticated request. weight financial systems, those in which Unlike settlements with physical tokens, the economic stakes or number of partic- questionable transactions can be reversed ipants or transactions is relatively low. In quickly and easily. such cases, confidentiality tends to be less The problem with traditional ledgers, of an issue; even if participants pay close however, is concentration of control. By attention to what the others are doing, putting so much power in one place we they won’t learn much of value. And it is create a significant security challenge, in precisely because the stakes are lower both technical and human terms. If some- that we might prefer to avoid the hassle one outside the process can hack into the and cost of setting up an intermediary. database, they can change the ledger at Some obvious examples of light- will and steal others’ funds, or destroy its weight financial systems include crowd- contents completely. Even worse, some- funding, gift cards, loyalty points, and one on the inside could corrupt the local currencies—and especially where ledger, and this kind of attack is harder to assets are redeemable in more than one detect or prove. As a result, wherever we place. We have also seen some blockchain use a centralized ledger, we must invest use in the mainstream financial sector, significant time and money in mecha- such as peer-to-peer trading between nisms that protect the ledger’s integrity. asset managers who are not in direct com- Many cases require ongoing verification petition. Blockchain is also being tested as using batch-based reconciliation between an internal accounting system in large the central ledger and those of each trans- organizations where each department or acting party. location must maintain control of its Enter the blockchain, or shared funds. In all of these cases, the lower cost ledger, which provides the benefits of and reduced friction of blockchains pro- ledgers without the problem of concen- vide immediate benefits, and the potential tration. Each entity instead runs a node loss of confidentiality is not a major con- that holds a copy of the ledger and main- cern. tains full control over its own assets, which are protected by private keys. Tracking Provenance Transactions propagate between nodes in a peer-to-peer fashion, while the A second class of usage we repeatedly blockchain ensures that consensus is hear about from MultiChain users is maintained. This architecture leaves no tracking the origin and movement of central attack point through which a high-value items across a supply chain, hacker or insider could corrupt the such as luxury goods, pharmaceuticals, ledger’s contents. As a result, a digital cosmetics, and electronics. Another is financial system can be deployed more critical documentation, such as bills of quickly and cheaply, with the added ben- lading or letters of credit. In supply chains efit of automatic reconciliation in real stretching across time and distance, all of time. these items suffer from counterfeiting and As discussed earlier, the downside is theft. that those using a shared ledger see all the The problem can be addressed using transactions taking place, which renders blockchains in the following way: when it unusable in situations where confiden- the high-value item is created, a corre- tiality is required. Blockchains are more sponding digital token is issued by a trust- suitable for what we might call light- ed entity, which then authenticates its point of origin. The digital token is moved

64 innovations / Blockchain for Global Development

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Where Blockchains Add Real Value

in parallel every time the physical item places, most tokens move in a single changes hands, thus the real-world chain direction, from origin to endpoint, with- of custody is precisely mirrored by a chain out being traded repeatedly back and of transactions on the blockchain. forth between the blockchain’s partici- If you like, the token acts as a virtual pants. If competitors rarely transact with certificate of authenticity, which is far each other (e.g., toy manufacturer to toy harder to steal or forge than a piece of manufacturer, or retailer to retailer), they paper. Upon receiving the digital token, cannot learn one another’s blockchain the final recipient of the physical item, “addresses” and connect them to real- whether a bank, distributor, retailer, or world identities. Furthermore, the activity customer, can verify the chain of custody can be easily partitioned into multiple all the way back to the point of origin. ledgers, each representing a different Indeed, in the case of documentation order or type of good. such as bills of lading, the physical item can be done away with altogether. INTERORGANIZATIONAL While all of this makes sense, the RECORD-KEEPING astute reader will notice that a regular database, managed (say) by an item’s Both of the previous use cases are based manufacturer, can accomplish the same on tokenized assets; that is, on-chain rep- task. This database can store a record of resentations of an item of value that is the current owner of each item, accept transferred between participants. signed transactions that represent each However, there is a second group of change of ownership, and respond to blockchain uses that is not related to incoming requests regarding the current assets. For this group the chain acts as a state of play. mechanism for collectively recording and So why use a blockchain instead? The notarizing any type of data, whose mean- answer is that, for this type of application, ing can be financial or otherwise. there’s a benefit to distributed trust. No One such example is an audit trail of matter where a centralized database is critical communications between two or held, there will be people in that place more organizations, say in the healthcare who have the ability (and can be bribed) or legal sectors. No individual organiza- to corrupt its contents—for example, by tion in the group can be trusted with marking forged or stolen items as legit. By maintaining this archive of records contrast, if provenance is tracked on a because falsified or deleted information blockchain that belongs collectively to a would significantly damage the others. supply chain’s participants, no individual Nonetheless, it is vital that all agree on the entity or small group of entities can cor- archive’s contents in order to prevent dis- rupt the chain of custody. As a bonus, dif- putes. ferent tokens (say, for some goods and the To solve this problem, the partici- corresponding bill of lading) can be pants need a shared database into which exchanged safely and directly, with a two- all the records are written, with each way swap guaranteed at the lowest record accompanied by a timestamp and blockchain level. proof of origin. The standard solution What about the problem of confiden- would be to create a trusted intermediary tiality? Blockchains’ suitability for supply whose role is to collect and store the chain provenance is a happy result of this records centrally. But blockchains offer a application’s simple pattern of transac- different approach, which gives the tions. In contrast to financial market- organizations a way to manage this

innovations / volume 12, number 1/2 65

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Gideon Greenspan

archive jointly while preventing individ- ing, because the entire purpose is to create ual participants (or small groups thereof) a shared archive that all participants can from corrupting it. see, even if some data is encrypted or This application of blockchains is hashed. Indeed, in some cases a actually not new at all. For 20 years, blockchain can help manage access to Z/Yen has been building systems in which confidential off-chain data by providing multiple entities collectively manage a an immutable record of digitally signed shared digital audit trail, using time- access requests. Either way, the straight- stamping, digital signatures, and a round- forward benefit of disintermediation is robin consensus scheme.5. While these that no additional entity must be created systems were not called blockchains, they and trusted to maintain this record. are technically identical in every respect. We might say that there is nothing new Multiparty Aggregation about using a blockchain for interorgani- zational recordkeeping—it’s just that the Technically speaking, this final class of world has finally become aware of the blockchain use is similar to the previous possibility. one, in that multiple parties are writing In terms of the actual data stored on data to a collectively managed record. the blockchain, there are three popular However, in this case the motivation is to options: overcome the infrastructural difficulty of combining information from a large 
Unencrypted data. This can be read by number of separate sources. every participant in the blockchain, Imagine two banks with internal which provides full collective trans- databases of customer identity verifica- parency and immediate resolution in tions. At some point they notice that they the case of a dispute. share a lot of customers, so they enter a 
Encrypted data. This can only be read reciprocal sharing arrangement in which by participants with the appropriate they exchange verification data to avoid decryption key. In the event of a dis- duplicated work. Technically, the agree- pute, anyone can reveal this key to a ment is implemented using standard trusted authority, such as a court, and master-slave data replication, in which use the blockchain to prove that the each bank maintains a live read-only copy original data was added by a certain of the other’s database and runs queries in party at a certain point in time. parallel against its own database and the 
Hashed data. A “hash” acts as a com- replica. So far, so good.6. pact digital fingerprint, which repre- Now imagine that these two banks sents a commitment to a particular invite three others to participate in this piece of data while keeping that data circle of sharing. Each of the five banks hidden. Any party that receives data runs its own master database, along with can easily confirm whether it matches a four read-only replicas of the others. With given hash, but inferring data from its five masters and 20 replicas, we have 25 hash is computationally impossible. total database instances. While doable, Only the hash is placed on the this consumes noticeable time and blockchain, while the original data is resources in each bank’s IT department. stored off-chain by interested parties, Fast-forward to the point where 20 who can reveal it in case of a dispute. banks are sharing information in this Naturally, confidentiality is not an way, and we’re looking at 400 total data- issue for interorganizational record-keep- base instances. For 100 banks, we reach

66 innovations / Blockchain for Global Development

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Where Blockchains Add Real Value

10,000 instances. In general, if every party Most importantly, in each case the is sharing information with every other, blockchain is addressing a real business the total number of database instances problem that could not be solved using a grows with the square of the number of regular database. participants. At some point in this process, the system is bound to break Workflow Management for down. Infrastructure Projects One obvious solution is for all the Construtivo is a Brazilian software com- banks to submit their data to a trusted pany that builds solutions for the design intermediary, whose job is to aggregate and construction phases of large infra- that data in a single master database. Each structure projects.9. For the past 15 years, bank could then query this database Construtivo’s general approach has been remotely, or run a local read-only replica to deliver software-as-a-service, which within its own four walls. While there’s means that the company acts as the cen- nothing wrong with this approach, tral trusted intermediary for managing blockchains offer a cheaper alternative, in project data. This is the traditional which the shared database is run directly approach to ensuring that all stakeholders by the banks that use it. Blockchains also maintain a consistent view of a project’s bring the added benefit of redundancy status and progress. and failover for the system as a whole. To satisfy their customers’ desire for It’s important to clarify that a greater transparency and auditability, blockchain does not act only as a distrib- Construtivo has now integrated a uted database such as Cassandra7. or blockchain into its solution, which pro- MongoDB.8. Unlike these systems, each vides the option of storing crucial project blockchain node enforces a set of rules data on-chain alongside Construtivo’s that prevents one participant from modi- database. Several infrastructure projects fying or deleting the data added by anoth- in South America are already using this er. There still appears to be some confu- option. Each project has its own chain, sion about this—indeed, one recently with nodes run by both Construtivo and released blockchain platform can be bro- stakeholders, such as contractors and ken by a single misbehaving node. In any engineering companies. Depending on event, a good platform will also make it the project’s requirements, the chain can easy to manage networks with thousands record plans, contracts, and other work- of nodes, which join and leave at will if flow-related information, and partici- granted the appropriate permissions. pants can browse through it using a web- based interface. BLOCKCHAINS IN The typical network for an infrastruc- PRODUCTION ture project has four nodes, with an aver- age transaction size of 15K. All nodes in Let’s conclude by reviewing several cases each chain participate in the validation in which permissioned blockchains run- process, while control over user permis- ning on our MultiChain platform have sions remains in Construtivo’s hands. been used in production, following the release of version 1.0 in the summer of Shared Ledger for a Catastrophe 2017. Each application described below Bond was independently built by a third party and is running in a network of four nodes Solidum Partners is an investment advi- or more, with multiple active validators. sory company that specializes in creating

innovations / volume 12, number 1/2 67

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Gideon Greenspan

catastrophe bonds.10. These are financial their own node, which gives them direct instruments that pay investors a higher control over their on-chain assets. While rate of yield than regular commercial a trustee knows the real-world identity bonds, but they carry the risk of only par- behind each address on the blockchain, tial or no repayment if a particular event participants do not know each other’s. occurs. In essence, purchasers of catastro- (Unlike many financial uses, the level of phe bonds are acting like insurance com- activity is not high enough for this veil of panies, providing the capital to cover confidentiality to be broken.) After com- unlikely losses and making a tidy profit as pleting anti-money-laundering and long as those losses don’t materialize. know-your-customer checks, Solidum In order to be easy to trade, nonphys- gives users access to the chain, and they ical securities like catastrophe bonds are can then transact with each other directly. traditionally held by a trusted intermedi- The network currently has around ten ary on their owners’ behalf. Trades in the nodes, four of which are permanently security are “settled” virtually via an online and participate in the consensus update of the intermediary’s records. For process. Solidum, the intermediary of choice was Euroclear, which holds over $30 trillion in Transaction Notarization for E- financial assets on behalf of investors, or Commerce more than 10 percent of the world’s Cryptologic, a blockchain consultancy total.11. Naturally, with 4,000 employees in based in Rosario, Argentina, has built and 15 offices around the world, Euroclear deployed a system for notarizing e-com- doesn’t provide this service for free. merce transactions as a way to help Due to recent changes at one of its resolve disputes between buyers and sell- banking partners, Solidum lost access to ers.12. Their first customer was Euroclear and had to seek another way to MercadoLibre, Latin America’s most manage the ledger. They issued a new $15 popular e-commerce site, which has million catastrophe bond directly onto a almost $1 billion in annual revenues.13. blockchain, along with dollar-denominat- Usually, when a customer makes a ed tokens that could be used for transact- purchase from an online merchant, they ing. If you like, they performed two pri- have to trust that merchant to record the vate-placement initial coin offerings, or transaction securely and permanently. In ICOs, but these had real underlying assets practice, however, nothing stops the mer- instead of a white paper and the hope of chant’s employees from deleting or mod- future value. ifying transaction records, which can cre- The blockchain enables safe delivery- ate a back door that delays delivery or lets versus-payment transactions, in which goods end up in the wrong hands. By con- two users exchange dollars and bond trast, if each transaction is recorded on a units in a single step—a feat that tradi- blockchain whose contents are publicly tionally requires help from a trusted visible and control of which is spread intermediary. Aside from avoiding this among a number of different parties, this middleman’s fees, using a permissioned record becomes far more difficult to blockchain gave Solidum easy and direct change retroactively. control over who can participate in the To preserve confidentiality, transac- system, and it did so without triggering tion data is hashed before being embed- the heavy regulation faced by Euroclear ded in the chain. The hashes provide a and its peers. mechanism for timestamping and nota- Each participant in the network has

68 innovations / Blockchain for Global Development

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021 Where Blockchains Add Real Value

rizing, and they are sufficient to settle cific cases, such as interbank settlement, later disputes if either party reveals the supply-chain finance, and shared identity, unhashed transaction. The network cur- but in reality blockchains can be applied rently contains seven permanent nodes whenever users seek to avoid having cen- that are spread between Cryptologic, var- tralized control over a digital system of ious government offices, and a partner record. abroad. Since the transactions contain only hashes, they are fairly small, and the 1.See http://www.ingentaconnect.com/con- network has seen a peak rate of 50 trans- tent/hsp/jpss/2016/00000010/00000002/art00 actions per second. 002?crawler=true&mimetype=application%2 52Fpdf&trendmd-shared=0 GENERAL LESSONS 2. See LEARNED https://www.multichain.com/blog/2015/09/de livery-versus-payment-blockchain/. We have provided some early examples of 3. See https://people.xiph.org/~greg/confiden- permissioned blockchains in production. tial_values.txt. The networks are still small, and their 4. modest transaction volume is far below See http://zerocash-project.org/paper. the limits of products like ours. 5. See http://www.zyen.com. Therefore, it’s important not to extrapo- 6. See late too much from these examples. https://en.wikipedia.org/wiki/Replication_(co Nonetheless, it’s interesting to note mputing)#Database_replication. what these applications have in common. 7. See http://cassandra.apache.org. First and most importantly, rather than using a blockchain for a blockchain’s 8. See https://www.rethinkdb.com. sake, they all derive from a genuine desire 9. See http://construtivo.com. for decentralization. All three cases 10. See http://solidumpartners.ch. demonstrate clear reasons to choose a blockchain architecture over messaging 11. See https://www.euroclear.com/. or a centralized database. 12. See https://cryptologic.io. Second, none of the chains has yet 13. See http://www.mercadolibre.com. transitioned to a decentralized model for administrator, who onboards new users and grants them permission to transact. It remains to be seen how often decentral- ized governance (as supported by MultiChain’s admin consensus model) is viable or necessary in practice. Perhaps it is sufficient for the blockchain to provide a transparent view of all administrators’ activity, while leaving control of this activ- ity with a single party. Finally, the nature of these applica- tions confirms our view that blockchains are a general-purpose technology for shared databases and are not restricted to particular industries. The lion’s share of media coverage might be received by spe-

innovations / volume 12, number 1/2 69

Downloaded from http://direct.mit.edu/itgg/article-pdf/12/1-2/58/705259/inov_a_00267.pdf by guest on 24 September 2021