One implementation detail that stood out to me in Newton Protocol's documentation is its two-phase consensus process before BLS signature aggregation.
Instead of having operators immediately evaluate and sign a task, Newton first separates data collection from policy evaluation. During the Prepare phase, operators independently fetch external data and return unsigned responses. The Gateway then computes canonical median values and verifies that every response falls within the configured tolerance. Only after this shared dataset is established does the Commit phase begin, where operators evaluate the policy using identical inputs and produce BLS signatures.
This sequence reduces the chance that operators sign different results because of small differences in time-sensitive data, such as external price feeds. The documentation also notes that responses outside the configured tolerance cause consensus to fail rather than being silently ignored.
Technical Question: How does the configured consensus tolerance affect the balance between network availability and consistency when external data changes rapidly@NewtonProtocol #newt $NEWT