Oracles and Ethereum

Oracles are a mechanism that allows applications to get information from the outside world that is not on-chain yet. One example is accessing the state of an Ethereum smart contract from a Linera application.

The contract runtime currently has three oracle methods:

• query_service allows the contract to make a call to the application's own service code. Services can access some off-chain information, so these are not guaranteed to return the same result each time they are called.
• http_post makes an HTTP POST request and returns the response.
• assert_before asserts that the block is being validated before a given time.

The first two are disabled on public devnets and testnets for now, but can be used locally by compiling with the unstable-oracles flag.

Applications should use these methods only in ways that make it very likely that all validators see the same result, otherwise any block proposals running that application's code are likely to fail, hurting the liveness of the users' chains. (Also, oracle methods are disallowed completely in fast rounds; see Chain Ownership Semantics.)

The Linera SDK uses http_post to implement the EthereumClient type, which provides functions to query an Ethereum node:

• get_balance for accessing the balance of an Ethereum account at a specific block number.

• read_events for reading events from a specified Ethereum smart contract from one block to the last block.

• non_executive_call for executing a function in an Ethereum smart contract at a specific block, with the result not being included in the chain.

None of those functions are changing the state of the Ethereum blockchain. They are all about observing the chain. All those functions take a specific block number in input since otherwise different validators could read different number of blocks.

The smart contract ethereum-tracker provides one example of such a contract and the end-to-end test test_wasm_end_to_end_ethereum_tracker how the interaction is actually done.