Flo: A Semantic Foundation for Progressive Stream Processing

Watch this 17-minute conference talk from POPL 2025 that introduces Flo, a parameterized streaming language providing a semantic foundation for progressive stream processing. Learn how researchers from UC Berkeley and Princeton University identify two key semantic properties—streaming progress and eager execution—that ensure deterministic outputs and fresh data in streaming systems. Discover how Flo abstracts over dataflow operators and stream structures while using a lightweight type system to distinguish between bounded and unbounded streams. The presentation demonstrates how Flo can model semantics from representative systems like Flink, LVars, and DBSP, showing the generality of its approach to unify various streaming language paradigms. This ACM SIGPLAN talk explores formal definitions that help answer the fundamental question of what makes a language truly "streaming" despite the variety of semantic models in existing systems.