FOCIL specs are being updated for Hegotá.
FOCIL for Hegotá
- FOCIL has been proposed for Hegotá. The CL headliner decision will be made in ACDC on the 19th.
Development Updates
-
The CL spec will have two main changes: IL bitlist inclusivity check and IL compliance check.
-
An IL bitlist is added to bids so builders can signal which ILs they considered when building a block. The bitlist serves as an informational signal for proposers on whether the corresponding payload is likely to satisfy IL constraints, but it is not a guarantee. For instance, a payload can be IL compliant when the bitlist is not inclusive, if ILs overlap to cover the missing IL. Therefore, the bitlist inclusivity is not used in fork choice. However, the bitlist information can be used out-of-protocol, e.g. to downscore builders that consistently provide non-inclusive bitlists or IL incompliant blocks.
-
When an attester for slot N receives an envelope for slot N-1, they check if the payload satisfies the IL constraints against their local IL view. If the payload is not IL compliant, the attester expects the beacon block for slot N to extend the envelope for slot N-2 rather than N-1, effectively reorging the IL incompliant envelope.
Q: If the bitlist is a superset of attesters’ merged view, could this require including transactions that not everybody has seen?
A: Builders have extra time to receive and update their IL view, so they may see more ILs than attesters. However, attesters validate IL compliance against their own local IL view, not the IL bitlist. A builder including more transactions than expected does not violate any FOCIL mechanism.
IL Analysis using Xatu Data
-
Pelle shared results from the IL analysis using Xatu data from PandaOps. With the current setup, roughly 20% of transactions in a mock IL are already included in the next block.
-
The data is noisy due to spam transactions and the inability to check state on historical data. Pelle is working with new data from Block Native for higher-resolution analysis and various IL building algorithms.