Data model

Description

This article covers the data model: how states, transactions and code contracts interact with each other and how they are represented in the code. It doesn't attempt to give detailed design rationales or information on future design elements: please refer to the R3 wiki for background information.

We begin with the idea of a global ledger. In our model, although the ledger is shared, it is not always the case that transactions and ledger entries are globally visible. In cases where a set of transactions stays within a small subgroup of users it should be possible to keep the relevant data purely within that group.

To ensure consistency in a global, shared system where not all data may be visible to all participants, we rely heavily on secure hashes like SHA-256 to identify things. The ledger is defined as a set of immutable states, which are created and destroyed by digitally signed transactions. Each transaction points to a set of states that it will consume/destroy, these are called inputs, and contains a set of new states that it will create, these are called outputs.

States contain arbitrary data, but they always contain at minimum a hash of the bytecode of a code contract, which is a program expressed in some byte code that runs sandboxed inside a virtual machine. Code contracts (or just "contracts" in the rest of this document) are globally shared pieces of business logic. Contracts define a verify function, which is a pure function given the entire transaction as input.

To be considered valid, the transaction must be accepted by the verify function of every contract pointed to by the input and output states. Beyond inputs and outputs, transactions may also contain commands, small data packets that the platform does not interpret itself, but which can parameterise execution of the contracts. They can be thought of as arguments to the verify function. Each command has a list of public keys associated with it. The platform ensures that the transaction is signed by every key listed in the commands before the contracts start to execute. Public keys may be random/identityless for privacy, or linked to a well known legal identity via a public key infrastructure (PKI).

Commands are always embedded inside a transaction. Sometimes, there's a larger piece of data that can be reused across many different transactions. For this use case, we have attachments. Every transaction can refer to zero or more attachments by hash. Attachments are always ZIP/JAR files, which may contain arbitrary content. Contract code can then access the attachments by opening them as a JarInputStream (this is temporary and will change later).

Note that there is nothing that explicitly binds together specific inputs, outputs, commands or attachments. Instead it's up to the contract code to interpret the pieces inside the transaction and ensure they fit together correctly. This is done to maximise flexibility for the contract developer.

Transactions may sometimes need to provide a contract with data from the outside world. Examples may include stock prices, facts about events or the statuses of legal entities (e.g. bankruptcy), and so on. The providers of such facts are called oracles and they provide facts to the ledger by signing transactions that contain commands they recognise, or by creating signed attachments. The commands contain the fact and the signature shows agreement to that fact. Time is also modelled as a fact, with the signature of a special kind of oracle called a timestamping authority (TSA). A TSA signs a transaction if a pre-defined timestamping command in it defines a after/before time window that includes "true time" (i.e. GPS time as calibrated to the US Naval Observatory). An oracle may prefer to generate a signed attachment if the fact it's creating is relatively static and may be referred to over and over again.

As the same terminology often crops up in different distributed ledger designs, let's compare this to other distributed ledger systems you may be familiar with. You can find more detailed design rationales for why the platform differs from existing systems in the R3 wiki, but to summarise, the driving factors are:

Improved contract flexibility vs Bitcoin
Improved scalability vs Ethereum, as well as ability to keep parts of the transaction graph private (yet still uniquely addressable)
No reliance on proof of work
Re-use of existing sandboxing virtual machines
Use of type safe GCd implementation languages.
Simplified auditing

Comparison with Bitcoin

Similarities:

The basic notion of immutable states that are consumed and created by transactions is the same.
The notion of transactions having multiple inputs and outputs is the same. Bitcoin sometimes refers to the ledger as the unspent transaction output set (UTXO set) as a result.
Like in Bitcoin, a contract is pure function. Contracts do not have storage or the ability to interact with anything. Given the same transaction, a contract's accept function always yields exactly the same result.
Bitcoin output scripts are parameterised by the input scripts in the spending transaction. This is somewhat similar to our notion of a command.
Bitcoin transactions, like ours, refer to the states they consume by using a (txhash, index) pair. The Bitcoin protocol calls these "outpoints". In our prototype code they are known as StateRefs but the concept is identical.
Bitcoin transactions have an associated timestamp (the time at which they are mined).

Differences:

A Bitcoin transaction has a single, rigid data format. A "state" in Bitcoin is always a (quantity of bitcoin, script) pair and cannot hold any other data. Some people have been known to try and hack around this limitation by embedding data in semi-standardised places in the contract code so the data can be extracted through pattern matching, but this is a poor approach. Our states can include arbitrary typed data.
A Bitcoin transaction's acceptance is controlled only by the contract code in the consumed input states. In practice this has proved limiting. Our transactions invoke not only input contracts but also the contracts of the outputs.
A Bitcoin script can only be given a fixed set of byte arrays as the input. This means there's no way for a contract to examine the structure of the entire transaction, which severely limits what contracts can do.
Our contracts are Turing-complete and can be written in any ordinary programming language that targets the JVM.
Our transactions and contracts have to get their time from an attached timestamp rather than a block chain. This is important given that we are currently considering block-free conflict resolution algorithms.
We use the term "contract" to refer to a bundle of business logic that may handle various different tasks, beyond transaction verification. For instance, currently our contracts also include code for creating valid transactions (this is often called "wallet code" in Bitcoin).

Comparison with Ethereum