
- Who
- AWS Machine Learning Blog authors
- What
- explain an open-source engine for comparing formal policies with semantic diffing
- Quote
The open source engine used in this post computes this diff using anti-unification , which finds a most specific common pattern for two expressions.
What happened
Ferhat Erata explains how an open source engine uses anti-unification over an e-graph to compare formal policies while accounting for declared algebraic laws and domain rewrites. Worked examples show how the method isolates disagreements such as a weakened authorization condition or a strict versus non-strict size threshold. The engine represents shared structure and preserves both alternatives at each difference for review.
The story so far
AWS Machine Learning Blog author Ferhat Erata explained how an open-source engine uses anti-unification over an e-graph to compare formal policies, showing where two policies differ while preserving both alternatives for review. Examples identify a weakened authorization condition and a strict versus non-strict size threshold. The comparison accounts for declared algebraic laws and domain rewrites. The engine does not evaluate Dogwood policies; that capability is not part of the current engine.
