Bastian Köcher
2675741a09
* Lol * Yeah * Moare * adaasda * Convert AURA to new pallet macro * AURA: Switch to `CurrentSlot` instead of `LastTimestamp` This switches AURA to use `CurrentSlot` instead of `LastTimestamp`. * Add missing file * Update frame/aura/src/migrations.rs Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com> * Remove the runtime side provide inherent code * Use correct weight * Add TODO * Remove the Inherent from AURA * 🤦 * Remove unused stuff * Update primitives authorship * Fix babe inherent data provider * Fix consensus-uncles * Fix BABE * Do some further changes to authorship primitives... :D * More work * Make it compile the happy path * Make it async! * Take hash * More stuff * Hacks * Revert "Hacks" This reverts commit cfffad88668cfdebf632a59c4fbfada001ef8251. * Fix * Make `execute_block` return the final block header * Move Aura digest stuff * Make it possible to disable equivocation checking * Fix fix fix * Some refactorings * Comment * Fixes fixes fixes * More cleanups * Some love * Better love * Make slot duration being exposed as `Duration` to the outside * Some slot info love * Add `build_aura_worker` utility function * Copy copy copy * Some stuff * Start fixing pow * Fix pow * Remove some bounds * More work * Make grandpa work * Make slots use `async_trait` * Introduce `SharedData` * Add test and fix bugs * Switch to `SharedData` * Make grandpa tests working * More Babe work * Make grandpa work * Introduce `SharedData` * Add test and fix bugs * Switch to `SharedData` * Make grandpa tests working * More Babe work * Make it async * Fix fix * Use `async_trait` in sc-consensus-slots This makes the code a little bit easier to read and also expresses that there can always only be one call at a time to `on_slot`. * Make grandpa tests compile * More Babe tests work * Fix network test * Start fixing service test * Finish service-test * Fix sc-consensus-aura * Fix fix fix * More fixes * Make everything compile *yeah* * Make manual-seal compile * More fixes * Start fixing Aura * Fix Aura tests * Fix Babe tests * Make everything compile * Move code around and switch to async_trait * Fix Babe * Docs docs docs * Move to FRAME * Fix fix fix * Make everything compile * Last cleanups * Fix integration test * Change slot usage of the timestamp * We really need to switch to `impl-trait-for-tuples` * Update primitives/inherents/src/lib.rs Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com> * Update primitives/inherents/src/lib.rs Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com> * Update primitives/inherents/src/lib.rs Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com> * Some extra logging * Remove dbg! * Update primitives/consensus/common/src/import_queue/basic_queue.rs Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com> Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
BABE (Blind Assignment for Blockchain Extension)
BABE is a slot-based block production mechanism which uses a VRF PRNG to randomly perform the slot allocation. On every slot, all the authorities generate a new random number with the VRF function and if it is lower than a given threshold (which is proportional to their weight/stake) they have a right to produce a block. The proof of the VRF function execution will be used by other peer to validate the legitimacy of the slot claim.
The engine is also responsible for collecting entropy on-chain which will be used to seed the given VRF PRNG. An epoch is a contiguous number of slots under which we will be using the same authority set. During an epoch all VRF outputs produced as a result of block production will be collected on an on-chain randomness pool. Epoch changes are announced one epoch in advance, i.e. when ending epoch N, we announce the parameters (randomness, authorities, etc.) for epoch N+2.
Since the slot assignment is randomized, it is possible that a slot is assigned to multiple validators in which case we will have a temporary fork, or that a slot is assigned to no validator in which case no block is produced. Which means that block times are not deterministic.
The protocol has a parameter c [0, 1] for which 1 - c is the probability
of a slot being empty. The choice of this parameter affects the security of
the protocol relating to maximum tolerable network delays.
In addition to the VRF-based slot assignment described above, which we will call primary slots, the engine also supports a deterministic secondary slot assignment. Primary slots take precedence over secondary slots, when authoring the node starts by trying to claim a primary slot and falls back to a secondary slot claim attempt. The secondary slot assignment is done by picking the authority at index:
blake2_256(epoch_randomness ++ slot_number) % authorities_len.
The secondary slots supports either a SecondaryPlain or SecondaryVRF
variant. Comparing with SecondaryPlain variant, the SecondaryVRF variant
generates an additional VRF output. The output is not included in beacon
randomness, but can be consumed by parachains.
The fork choice rule is weight-based, where weight equals the number of primary blocks in the chain. We will pick the heaviest chain (more primary blocks) and will go with the longest one in case of a tie.
An in-depth description and analysis of the protocol can be found here: https://research.web3.foundation/en/latest/polkadot/block-production/Babe.html
License: GPL-3.0-or-later WITH Classpath-exception-2.0