BigchainDB 2.

0
The Blockchain Database
BigchainDB GmbH, Berlin, Germany
May 2018
Paper version 1.0

Abstract
BigchainDB is software that has blockchain properties (e.g. decentral-
ization, immutability, owner-controlled assets) and database properties
(e.g. high transaction rate, low latency, indexing & querying of structured
data). It was first released—open source—in February 2016 and has been
improved continuously ever since. BigchainDB version 2.0 makes signifi-
cant improvements over previous versions. In particular, it is now Byzan-
tine fault tolerant (BFT), so up to a third of the nodes can fail in any way,
and the system will continue to agree on how to proceed. BigchainDB 2.0
is also production-ready for many use cases. In this paper, we review the
design goals of BigchainDB 2.0 and how they were achieved, we explore
some use cases, we show how BigchainDB fits into the overall decentral-
ization ecosystem, we follow the life of a transaction to understand how
BigchainDB 2.0 works, we note ways to try BigchainDB, we outline how
you can contribute, and we summarize future plans.

1 BigchainDB 2.0 Design Goals

BigchainDB was first announced in February 2016 [1], along with the orig-
inal whitepaper and the first software release (version 0.1). BigchainDB is
called a blockchain database because it has some blockchain properties and
some database properties. The original design started with a database and
added some blockchain characteristics such as decentralization, immutability,
and owner-controlled assets. The idea was that the resulting system would
inherit the desirable properties of the database such as low latency, high trans-
action rate and high capacity.
It worked, mostly, but there were some issues. One issue was that the
resulting system couldn’t withstand arbitrary faults (in a bounded subset of the
nodes); it wasn’t Byzantine fault tolerant (BFT) [2]. Another issue was that
the underlying database has a primary (or master) node, and that node does
all the writes. The other nodes just replicate whatever the primary writes. The
primary node could be the primary for a long time. It was a single point of
control and a single point of failure. Another issue was that there was only one
logical database, so if a malicious agent managed to get admin privileges, they
could delete the entire database (across all the nodes) with a single command.

1
 BigchainDB 2.0 was designed to resolve all those issues while retaining the
desirable database and blockchain properties. Below, we go through each design
goal of BigchainDB 2.0 and outline how it was achieved.

Typical Distributed
Typical Blockchain BigchainDB
Database
Decentralization X X
Byzantine Fault
X X
Tolerance
Immutability X X
Owner-Controlled
X X
Assets
High Transaction
X X
Rate
Low Latency X X
Indexing
& Querying of X X
Structured Data

1.1 Full Decentralization and Byzantine Fault Tolerance

BigchainDB 2.0 uses Tendermint [3, 4] for all networking and consensus. Each
node has its own local MongoDB database [5], and all communication between
nodes is done using Tendermint protocols. One consequence is that the resulting
system is BFT, because Tendermint is BFT. Another result is that if a malicious
hacker manages to get admin privileges to one of the local MongoDB databases,
then the worst they can do is corrupt or delete the data in that local database;
the MongoDB databases in the other nodes won’t be affected. Tendermint still
has something like a primary node (the current block proposer), but it changes
with each round (using round robin) and the Tendermint developer team has
an open issue to make that more secure (e.g. less predictable) [6].
If every node in a BigchainDB 2.0 network is owned and operated by a
different person or entity, then it’s a decentralized network because it has no
single owner, no single point of control, and no single point of failure. Ideally,
the nodes should be located in many countries, legal jurisdictions and hosting
providers, so an issue with one doesn’t affect them all. Any node can fail and
the rest of the network will continue to operate. In fact, up to one third of the
nodes can fail1 in arbitrary ways, and the rest of the network will continue to
work, i.e. the non-faulty nodes will agree on how to proceed.

1.2 Immutability
Once data gets stored in a BigchainDB network, it can’t be changed or erased,
or at least not without great difficulty. If some data somehow manages to get
changed or erased, then that is detectable. One might say that the storage
1 Technically, up to one third of the voting power can fail, but we usually run BigchainDB

networks with all nodes having the same voting power, so one third of voting power is the
same as one third of the nodes.

2
is “practically immutable” but in the blockchain world, one usually just says
“immutable.”
BigchainDB uses several strategies to achieve practical immutability. The
simplest one is that there are no BigchainDB-provided APIs to change or erase
stored data.
Another strategy is that every node has a full copy of all the data in a stand-
alone MongoDB database (i.e. there is no global MongoDB database). Even if
one node gets corrupted or destroyed, the other nodes won’t be affected and
will still have a copy of all the data.
Another strategy is that all transactions are cryptographically signed. After
a transaction is stored, changing its contents will change the signature, which
can be detected (unless the public key is also changed, but that should also be
detectable because every block of transactions is signed by the a node, and the
public keys of all the nodes are all known).

1.3 Owner-Controlled Assets

Like most blockchains, BigchainDB has a concept of owner-controlled assets.
Only the owner (or owners) of an asset can transfer that asset. (The owners are
the holders of a particular set of private keys.) Not even a node operator can
transfer an asset.
In most blockchains, there’s only one built-in asset (e.g. Bitcoin or Ether),
but BigchainDB allows external users to create as many assets as they need.
However, it’s worth noting that a user can’t create assets that appear to be
created by someone else. All assets created by Mike are cryptographically signed
by Mike. For example, a user named Joe might decide to create 1000 “Joe
tokens.” He would do that by building a BigchainDB CREATE transaction,
signing it with his private key, and sending it to a BigchainDB network. (Later
in this paper, we trace what happens to a transaction when it arrives at a
BigchainDB network.) Initially, Joe might own all 1000 tokens, so only he
could transfer them to others. He could transfer 37 tokens to Lisa by creating
a BigchainDB TRANSFER transaction with two outputs: one with an amount
of 37 tokens with a condition that only Lisa can transfer it, and the other with
all remaining tokens (1000 − 37 = 963 tokens) with a condition that only Joe
can transfer it.2
BigchainDB checks every transaction to make sure it’s not trying to transfer
an output that was already transferred (spent) by another transaction, i.e. it
prevents double-spending. It also checks many other things, all of which are
listed in the BigchainDB Transaction Spec v2 [7].

1.4 High Transaction Rate

One design goal of BigchainDB has always been the ability to process a large
number of transactions each second. That’s still true with BigchainDB 2.0.
BigchainDB 2.0 was still in Alpha at the time of writing. Performance
tests were being written and started, but no concrete results were available
yet. (The final, stable BigchainDB 2.0 is expected in June of 2018.) However,
2 BigchainDB allows for quite complex conditions on the outputs. For example, an output

could have a condition which says that “Jamie OR (Pat AND Kelly) OR (three of {Casey,
Drew, Laurie, Riley or Whitney})” must sign.

3
since BigchainDB 2.0 is based on Tendermint, we can look at other Tendermint-
based networks to get a sense of what can be expected. According to the Cosmos
whitepaper [8]:

Despite its strong guarantees, Tendermint provides exceptional per-

formance. In benchmarks of 64 nodes distributed across 7 data-
centers on 5 continents, on commodity cloud instances, Tendermint
consensus can process thousands of transactions per second, with
commit latencies on the order of one to two seconds. Notably, per-
formance of well over a thousand transactions per second is main-
tained even in harsh adversarial conditions, with validators crashing
or broadcasting maliciously crafted votes.

1.5 Low Latency and Fast Finality

Tendermint-based networks (such as BigchainDB networks) take only a few
seconds (or less) for a transaction to be included in a new committed block.
Once that happens, there’s no way it can be reverted or considered defunct in
the future, because Tendermint doesn’t do forking.

1.6 Indexing & Querying Structured Data

Each node in a BigchainDB 2.0 network has its own local MongoDB database.
That means that each node operator has access to the full power of MongoDB
for indexing and querying the stored data (transactions, assets, metadata and
blocks, all of which are JSON strings). Each node operator is free to decide
how much of that power they expose to external users. One node operator
might decide to index geospatial data and offer optimized geospatial queries via
a REST API, whereas another node operator might decide to offer a GraphQL
API.
By default, BigchainDB 2.0 creates some MongoDB indexes and the Big-
chainDB HTTP API includes some endpoints for doing basic queries. However,
as outlined in the previous paragraph, each node operator can add additional
indexes and query APIs.

1.7 Sybil Tolerance

Some blockchain networks (such as Bitcoin) allow anyone to add their node to
the network. That brings the concern that someone could add so many nodes
that they effectively control the network: a Sybil attack [9]. Bitcoin makes
Sybil attacks unlikely by making them prohibitively expensive. In a BigchainDB
network, the governing organization behind the network controls the member
list, so Sybil attacks are not an issue.

2 Use Cases
BigchainDB, with the capabilities mentioned in the previous section, can serve a
multitude of use cases. Wherever there is a need of immutable, tamper-resistant
data representing digital assets, BigchainDB can be used. There are several
industry verticals which can directly benefit from BigchainDB features. In the

4
following sub-sections, we will briefly touch some of these industry verticals and
discuss how BigchainDB can be of help in these scenarios.

2.1 Supply Chain

In a typical supply chain scenario, there are several parties/entities collaborat-
ing and exchanging information with each other. Primarily, this information
is about the processes of tracking goods being manufactured until they reach
the logical end of the supply chain cycle (retailers/end-consumers). The major
challenge faced by these entities collaborating in a supply chain scenario is man-
agement and security of the information being shared. Eventually, several silos
of data emerge and it becomes hard to manage. That’s where blockchain tech-
nology, in general, can help organizing this data in a shared system so that the
overall management of information becomes easy. Because of immutability and
tamper resistance, blockchain also provides a layer of trust to the collaborating
entities so that they can trust the data even when they don’t trust each other.
While other blockchains and decentralized systems can help organize data in a
shared system,they are not optimized for the high throughput and query capa-
bilities required for a supply chain scenario. That’s where BigchainDB shines,
bringing along it’s query capabilities and high throughput performance. When
used in a supply chain scenario, like a regular blockchain, BigchainDB helps
organize data in a decentralized system, but, in addition, allows the users to
query the data to generate reports and do calculations on the fly.

2.2 Intellectual Property Rights Management

When it comes to IP rights management and provenance, blockchain technology
has several benefits as it helps provide immutability to artists’ claims. Once
an art asset is registered on a blockchain with proper attribution, it can be
used to prove the ownership of the IP rights. Ownership transfers can also be
recorded. For this reason, we created ascribe [10] in 2014 and it was based on
the Bitcoin blockchain. However, soon the throughput of the Bitcoin public
blockchain became a bottleneck. That’s when we envisioned BigchainDB and
began building it. BigchainDB 2.0 also handles high throughput and hence is
an excellent choice for IP use-cases.

2.3 Digital Twins and IoT

Digital twins are the digital representation of physical objects which can be
tracked on the basis of verification of authenticity and provenance, proof of
ownership, life-cycle traceability, and input data from IoT devices & sensors.
To manage this scale of information, giving every product and object a story
of its own, we need a high-throughput tamper-resistant system which can also
serve results quickly. That’s BigchainDB.

2.4 Identity
Identity is one of the most critical pieces when it comes to managing user-specific
information. It has become ever more important in scenarios like IoT and Digital
Twins where even machines have an identity. With identity theft becoming one

5
of the major concerns of today, we need to make sure that identity of a human
or a machine is self-sovereign and hack-proof. Because of a large amount of
data associated with identities, we also need to make sure that the systems
handling identity-related data are capable of handling high scale. That’s when
BigchainDB becomes a natural fit for solving identity-related use cases because
of its combined characteristics of blockchains and databases.

2.5 Data Governance

Data governance use cases are scenarios where multiple independent organiza-
tions or entities collaborate on defining common processes and directives. To-
day, the major challenge with data governance is lack of collaboration and trust.
Also, there are no clear incentives for the participants to collaborate on topics
of governance. BigchainDB helps solve these challenges in data governance by
providing an incentive-driven, easy-to-integrate platform. The approach is to
model data governance topics, feedback and economic incentives as BigchainDB
assets. Once we model a data governance system as a collection of BigchainDB
assets, it becomes fairly straightforward to collaborate and define processes on
top of them as the data is shared using a common substrate and all participants
can be incentivised to participate.

2.6 Immutable Audit Trails

Immutable audit trails are one of the generic use cases of BigchainDB. While
not directly linked with any industry vertical, they helps solve a lot of track and
trace challenges across verticals. From banking to supply chain, from utilities to
access control, audit trails are heavily relied upon. It really adds a lot of value
if these audit trails are immutable and easy to query. BigchainDB, because
of high throughput and query capability becomes a natural fit for maintaining
immutable audit trails.

2.7 Some Closing Remarks on Use Cases

In general, BigchainDB can be used in almost all the scenarios where there is a
need of immutable and tamper-proof storage of data assets at high throughput
with ability to search and query. BigchainDB can also be used by groups of
people or organizations who want to have a shared database, even if they don’t
trust (or know) each other. One must be careful about the data one stores in a
BigchainDB network. For example, storing personally-identifiable information
(PII) is discouraged, because many countries have regulations that require PII
to be erasable upon request.

3 BigchainDB in the Decentralization Ecosystem

As a blockchain database, BigchainDB is complementary to other decentral-
ized systems, such as decentralized file storage (e.g. IPFS [11]), decentralized
data exchange protocols (e.g. Ocean Protocol [12]), smart-contract blockchains
(e.g. Ethereum [13, 14] or Hyperledger Fabric [15]), and decentralized process-
ing (e.g. TrueBit [16]). BigchainDB works with centralized computing systems

6
as well. Figure 1 illustrates some ways BigchainDB could be used in various
technology stacks.

Centralized Decentralized

Applications Applications Applications

Platform (e.g. AWS) Platform (e.g. AWS) Platform (e.g. Ethereum)

Processing (e.g. EC2) Processing (e.g. EC2) Processing (e.g. Eth VM)

File System Database File System Database File System Database

(e.g. S3) (e.g. MySQL) (e.g. S3) BigchainDB (e.g. IPFS) BigchainDB

Figure 1: From a base context of a centralized cloud computing ecosystem (left),

BigchainDB can be added as another database to gain some decentralization
benefits (middle). It also fits into a fully-decentralized technology stack (right).

4 The Life of a BigchainDB Transaction, or How

BigchainDB 2.0 Works
One can get a good sense of how BigchainDB 2.0 works by following the life of
a BigchainDB transaction.

4.1 BigchainDB Transactions

A BigchainDB transaction is a JSON string that conforms to a BigchainDB
Transactions Specification (Spec). At the time of writing, there were two such
specs: v1 and v2. Transactions conforming to BigchainDB Transactions Spec
v1 were accepted by BigchainDB versions 1.0–1.3; such transactions are no
longer supported. Transactions conforming to BigchainDB Transactions Spec
v2 [17] are accepted by BigchainDB version 2.0 (i.e. the latest version at the
time of writing). Each transactions spec explains the expected keys and values
(including what they mean), instructions for how to construct a transaction, a
list of checks that must be done to check if a transaction is valid, and details
of the cryptographic primitives used. Listing 1 shows an example BigchainDB
transaction (v2).
If someone wants to construct a valid BigchainDB transaction, then they’ll
typically use a BigchainDB driver (software package). There’s a list of Big-
chainDB drivers in the BigchainDB docs [18].

7
{
"id": "3667c0e5cbf1fd3398e375dc24f47206cc52d53d771ac68ce14d c
,→ df0fde806a1c",
"version": "2.0",
"inputs": [
{
"fulfillment": "pGSAIEGwaKW1LibaZXx7_NZ5-V0alDLvrguGLyL c
,→ RkgmKWG73gUBJ2Wpnab0Y-4i-kSGFa_VxxYCcctpT8D6s4uTGOO c
,→ F-hVR2VbbxS35NiDrwUJXYCHSH2IALYUoUZ6529Qbe2g4G",
"fulfills": null,
"owners_before": [
"5RRWzmZBKPM84o63dppAttCpXG3wqYqL5niwNS1XBFyY"
]
}
],
"outputs": [
{
"amount": "1",
"condition": {
"details": {
"public_key":
,→ "5RRWzmZBKPM84o63dppAttCpXG3wqYqL5niwNS1XBFyY",
"type": "ed25519-sha-256"
},
"uri": "ni:///sha-256;d-_huQ-eG-QQD-GAJpvrSsy7lLJqyNh c
,→ tUAs_own7aTY?fpt=ed25519-sha-256&cost=131072"
},
"public_keys": [
"5RRWzmZBKPM84o63dppAttCpXG3wqYqL5niwNS1XBFyY"
]
}
],
"operation": "CREATE",
"asset": {
"data": {
"message": "Greetings from Berlin!"
}
},
"metadata": null
}

Listing 1: An Example BigchainDB Transaction (v2)

8
4.2 Sending a Transaction to a BigchainDB Network
Once one has a transaction, they can send it to a BigchainDB network using
the BigchainDB HTTP API [19]. More specifically, one would use one of the
following endpoints, with the transaction in the body of the HTTP request:

POST /api/v1/transactions
POST /api/v1/transactions?mode=async
POST /api/v1/transactions?mode=sync
POST /api/v1/transactions?mode=commit

Later, we’ll see what the different modes mean. A BigchainDB driver could
also be used to post a transaction. The HTTP request (containing the trans-
action) can be sent to any of the nodes in the BigchainDB network, or even
more than one. Figure 2 illustrates the main components in a four-node Big-
chainDB 2.0 network, and how they communicate with each other.
1

2
e

e
od

od
N

BigchainDB Node

mongoDB

BigchainDB Server
3

4
e

e
od

od

Tendermint
N

Figure 2: Communications in a Four-Node BigchainDB 2.0 Network

4.3 Arrival of a Transaction at a Node

The details of what happens next can vary. Let’s skip those details and assume
that the HTTP request arrives successfully at the Gunicorn web server inside a
BigchainDB node, because that’s where all incoming HTTP requests should get
routed. Gunicorn exposes a standard interface (Web Server Gateway Interface
[WSGI]) which enables Python applications to talk to it. (WSGI is a Python
standard; the spec is in Python Enhancement Proposal 3333 [20].) BigchainDB
uses the Flask web application development framework to simplify working with
WSGI/Gunicorn.
Flask is used to route the request to a Python method for handling that
endpoint. That method checks the validity of the transaction. If it’s not valid,

9
then that’s the end of the story for the transaction, the HTTP response status
code is 400 (Error), and the response body gives some information about what
was invalid. If the transaction is valid, then it’s converted to Base64 and put
into a new JSON string with some other information (such as the mode). Big-
chainDB then sends that string to the local Tendermint instance in the body of
an HTTP POST request. That request uses the Tendermint Broadcast API [21].
(Tendermint has other APIs.)

4.4 Arrival of a Transaction at a Tendermint Instance

To learn what happens to the transaction once it arrives in the local Tendermint
instance, one should read the Tendermint docs about the Broadcast API [21].
At the time of writing, those docs said:

When a transaction is sent to a Tendermint node, it will run via

CheckTx against the application. If it passes CheckTx, it will be
included in the mempool, broadcast to other peers, and eventually
included in a block.
Since there are multiple phases to processing a transaction, we offer
multiple endpoints to broadcast a transaction:
/broadcast tx async
/broadcast tx sync
/broadcast tx commit
These correspond to no-processing, processing through the mem-
pool, and processing through a block, respectively. That is,
broadcast tx async, will return right away without waiting to hear
if the transaction is even valid, while broadcast tx sync will return
with the result of running the transaction through CheckTx. Using
broadcast tx commit will wait until the transaction is committed
in a block or until some timeout is reached, but will return right
away if the transaction does not pass CheckTx. . .
The benefit of using broadcast tx commit is that the request re-
turns after the transaction is committed (i.e. included in a block),
but that can take on the order of a second. For a quick result, use
broadcast tx sync, but the transaction will not be committed until
later, and by that point its effect on the state may change.

The above text requires some explanation:

• CheckTx is an API that Tendermint expects BigchainDB to implement.

It’s explained below.
• If someone uses BigchainDB’s POST /api/v1/transactions?mode=async
endpoint to send the transaction, then the Tendermint
/broadcast tx async endpoint will be used. Similar things can be said
for the sync and commit modes. If no mode was specified, then the default
is async.

10
 • Every Tendermint instance has a local mempool (memory pool) of trans-
actions which have passed initial validation, but haven’t been included in
a block yet.

When Tendermint wants to determine if a transaction is valid, it sends the

transaction to BigchainDB using a CheckTx request. It expects BigchainDB to
implement CheckTx, and several other message types, all of which are explained
in the ABCI Specification [22]. (ABCI stands for Application BlockChain In-
terface.) In particular, BigchainDB implements:

• InitChain
• Info
• CheckTx
• BeginBlock
• DeliverTx
• EndBlock
• Commit

Tendermint takes care of proposing new blocks (each of which contains a

set of transactions) and making sure that all the nodes agree on the next
block in a Byzantine fault tolerant way. Each BigchainDB instance keeps
track of the block-under-construction by collecting the transactions delivered
(by DeliverTx) between the BeginBlock and EndBlock calls.
When Tendermint delivers a transaction to a BigchainDB instance using
DeliverTx, BigchainDB checks the validity of the transaction again. If it’s
valid, it keeps the transaction around (in memory), but it doesn’t write anything
to MongoDB yet. BigchainDB waits for the Commit message before writing the
new block (and all contained transactions) to MongoDB.
When storing a transaction in MongoDB, BigchainDB removes the asset.data
value (JSON string) and stores it in a separate MongoDB collection of assets.
It does that to facilitate the text search of assets. Similarly, the metadata value
(JSON string) is also removed and stored in a separate collection.
Tendermint writes the block to the blockchain (stored in a local LevelDB
database) after getting a response to the Commit message.
We glossed over some details above, because our goal was to give a big-
picture overview. If you want all the details, then you’re in luck! All the code
in question is open source, so you can look it up and read exactly what it does.

5 Further Reading (Technical Reading)

• BigchainDB Transactions Spec v2 [7]
• BigchainDB HTTP API [19]
• BigchainDB Enhancement Proposals [23]
• Tendermint Documentation [24]

11
 • Tendermint ABCI Specification [25]
• Papers about Tendermint found by Google Scholar [26]

6 How to Try BigchainDB

If you want to try BigchainDB, then you need a BigchainDB network to connect
to. You could deploy your own network, or you could use the BigchainDB
Testnet, a live BigchainDB network operated by the BigchainDB development
team.
If you go to the Get Started page on the BigchainDB website [27], you can
enter some text and click “Off you go.” An in-page JavaScript app will build
a BigchainDB transaction and send it to the BigchainDB Testnet. You can see
the constructed transaction and check if it was actually stored.
You can also write your own app to write to and read from a BigchainDB
network. It can be written using any programming language, but there are
official BigchainDB drivers (software packages) for JavaScript and Python, so
we recommend using one of those. Links can be found in the Drivers & Tools
page of the docs [18]. The docs for those drivers include example code.
Once you’ve written an app, you could test it against the BigchainDB Test-
net, but first you’ll need access credentials. You can get those (for free) by
signing up for an account at testnet.bigchaindb.com.

7 How to Contribute Ideas or Code

We changed our processes to make it easier for anyone to contribute ideas or
code to BigchainDB. Questions can be asked in one of the BigchainDB chat
rooms on Gitter [28]. Bug reports can be submitted by creating a new issue in
the relevant GitHub repository. Changes to code can be made by submitting
a pull request to the relevant GitHub repository. We use a variation of the
Collective Code Construction Contract (C4) [29] to guide how we handle pull
requests. Detailed feature requests can be made by submitting a BigchainDB
Enhancement Proposal (BEP) as a pull request in the bigchaindb/BEPs repos-
itory [30]. The format of a BEP should follow the outline given in our variant
of the Consensus-Oriented Specification System (COSS) [31].

8 The BigchainDB Roadmap

We aim to release the final, stable BigchainDB 2.0 in June of 2018. No new
features will be added before then; we’re mostly doing testing (of all kinds).
We’ll fix any issues that arise and report the results of performance tests once
we have them.
Starting with BigchainDB 2.0, there will always be tools and documenta-
tion to help with migration (including data migration) to future versions. Big-
chainDB 2.0 is production-ready for many uses cases.
There’s an online BigchainDB Roadmap [32] that lists our medium-term
goals. It’s updated fairly often. One goal is to create some example projects
illustrating how one can use BigchainDB with other decentralized systems, such

12
as Ethereum. Longer-term, we want to make it possible to create a public
BigchainDB network where anyone can add a node without getting permission.

9 Conclusion
BigchainDB is a blockchain database: it has both blockchain properties and
database properties. That combination makes it useful for a wide variety of
use cases, including supply chain, IP rights management, digital twins & IoT,
identity, data governance and immutable audit trails. BigchainDB 2.0 includes
significant improvements, including the integration of Tendermint for inter-node
networking and Byzantine fault tolerant (BFT) consensus. BigchainDB 2.0
is now production-ready for many uses cases. For more information, see the
BigchainDB website at bigchaindb.com.

References
[1] Carlo Thomas. ascribe announces scalable blockchain database Big-
chainDB. CoinReport, February 2016. https://coinreport.net/
ascribe-announces-scalable-blockchain-database-bigchaindb/.

[2] Leslie Lamport, Robert Shostak, and Marshall Pease. The Byzantine Gen-
erals Problem. ACM Transactions on Programming Languages and Systems
(TOPLAS), 4(3):382–401, July 1982. http://research.microsoft.com/
en-us/um/people/lamport/pubs/byz.pdf.
[3] Jae Kwon. Tendermint: Consensus without Mining, fall 2014.

[4] Tendermint. https://tendermint.com/.

[5] MongoDB. https://www.mongodb.com.
[6] Anton Kaliaev (melekes on GitHub). tendermint/tendermint Issue #763:
Introduce randomness into proposer selection? https://github.com/
tendermint/tendermint/issues/763.
[7] BigchainDB Transactions Spec v2. https://github.com/bigchaindb/
BEPs/tree/master/13.
[8] Cosmos: A Network of Distributed Ledgers. https://cosmos.network/
resources/whitepaper.
[9] John R Douceur. The Sybil Attack. In Peer-to-peer Systems, pages
251–260. Springer, 2002. http://research.microsoft.com/pubs/74220/
IPTPS2002.pdf.
[10] ascribe. https://www.ascribe.io/.

[11] J. Benet. IPFS – Content Addressed, Versioned, P2P File System. http:
//static.benet.ai/t/ipfs.pdf, 2014.
[12] Ocean Protocol: A Decentralized Substrate for AI Data & Services. https:
//oceanprotocol.com/tech-whitepaper.pdf.

13
[13] Ethereum. https://ethereum.org/.
[14] Vitalik Buterin. Ethereum White Paper: A Next Genera-
tion Smart Contract & Decentralized Application Platform.
http://blog.lavoiedubitcoin.info/public/Bibliotheque/
EthereumWhitePaper.pdf.
[15] Hyperledger Fabric. https://www.hyperledger.org/projects/fabric.
[16] Jason Teutsch and Christian Reitwießner. A scalable verification solution
for blockchains. November 2017.
[17] BEP-13: BigchainDB Transactions Spec v2. https://github.com/
bigchaindb/BEPs/tree/master/13.
[18] BigchainDB Drivers & Tools. https://docs.bigchaindb.com/projects/
server/en/latest/drivers-clients/index.html.
[19] BigchainDB HTTP API. https://docs.bigchaindb.com/projects/
server/en/latest/http-client-server-api.html.
[20] PEP 3333 – Python Web Server Gateway Interface v1.0.1. https://www.
python.org/dev/peps/pep-3333/.
[21] Tendermint Broadcast API. http://tendermint.readthedocs.io/
projects/tools/en/master/using-tendermint.html#broadcast-api.
[22] Tendermint ABCI Specification. https://github.com/tendermint/
abci/blob/master/specification.rst.
[23] BigchainDB Enhancement Proposals. https://github.com/bigchaindb/
BEPs.
[24] Tendermint Documentation. http://tendermint.readthedocs.io/
projects/tools/en/master/index.html.
[25] ABCI Overview. http://tendermint.readthedocs.io/projects/
tools/en/master/introduction.html#abci-overview.
[26] Papers about Tendermint found by Google Scholar. https://scholar.
google.de/scholar?hl=en&as_sdt=0%2C5&q=Tendermint&btnG=.
[27] BigchainDB – Get Started. https://www.bigchaindb.com/developers/
getstarted/.
[28] BigchainDB chat rooms. https://gitter.im/bigchaindb/home.
[29] BEP-1: The BigchainDB Variant of the Collective Code Construction Con-
tract (C4). https://github.com/bigchaindb/BEPs/tree/master/1.
[30] The bigchaindb/BEPs Repository on Github. https://github.com/
bigchaindb/BEPs.
[31] BEP-2: The BigchainDB Variant of the Consensus-Oriented Specification
System. https://github.com/bigchaindb/BEPs/tree/master/2.
[32] BigchainDB Roadmap. https://github.com/bigchaindb/org/blob/
master/ROADMAP.md.

14