- A valid transaction must be accepted by the contract of each of its input and output states
- Contracts are written in a JVM programming language (e.g. Java or Kotlin)
- Contract execution is deterministic and its acceptance of a transaction is based on the transaction’s contents alone
Recall that a transaction is only valid if it is digitally signed by all required signers. However, even if a transaction gathers all the required signatures, it is only valid if it is also contractually valid.
Contract validity is defined as follows:
- Each transaction state specifies a contract type
- A contract takes a transaction as input, and states whether the transaction is considered valid based on the contract’s rules
- A transaction is only valid if the contract of every input state and every output state considers it to be valid
We can picture this situation as follows:
The contract code can be written in any JVM language, and has access to the full capabilities of the language, including:
- Checking the number of inputs, outputs, commands, time-window, and/or attachments
- Checking the contents of any of these components
- Looping constructs, variable assignment, function calls, helper methods, etc.
- Grouping similar states to validate them as a group (e.g. imposing a rule on the combined value of all the cash states)
A transaction that is not contractually valid is not a valid proposal to update the ledger, and thus can never be committed to the ledger. In this way, contracts impose rules on the evolution of states over time that are independent of the willingness of the required signers to sign a given transaction.
The contract sandbox
Transaction verification must be deterministic - a contract should either always accept or always reject a given transaction. For example, transaction validity cannot depend on the time at which validation is conducted, or the amount of information the peer running the contract holds. This is a necessary condition to ensure that all peers on the network reach consensus regarding the validity of a given ledger update.
Future versions of Corda will evaluate transactions in a strictly deterministic sandbox. The sandbox has a whitelist that prevents the contract from importing libraries that could be a source of non-determinism. This includes libraries that provide the current time, random number generators, libraries that provide filesystem access or networking libraries, for example. Ultimately, the only information available to the contract when verifying the transaction is the information included in the transaction itself.
Developers can pre-verify their CorDapps are determinsitic by linking their CorDapps against the deterministic modules.
Since a contract has no access to information from the outside world, it can only check the transaction for internal validity. It cannot check, for example, that the transaction is in accordance with what was originally agreed with the counterparties.
Peers should therefore check the contents of a transaction before signing it, even if the transaction is contractually valid, to see whether they agree with the proposed ledger update. A peer is under no obligation to sign a transaction just because it is contractually valid. For example, they may be unwilling to take on a loan that is too large, or may disagree on the amount of cash offered for an asset.
Sometimes, transaction validity will depend on some external piece of information, such as an exchange rate. In these cases, an oracle is required. See Oracles for further details.
Each contract also refers to a legal prose document that states the rules governing the evolution of the state over time in a way that is compatible with traditional legal systems. This document can be relied upon in the case of legal disputes.