• Category: Dev tools
  • Status: discussion
  • Sources: future-file-format/f3, discussion
  • Summary: F3 ("Future File Format") is a columnar data file format from a SIGMOD 2026 paper by Xinyu Zeng, Wes McKinney, Jignesh Patel, Andrew Pavlo, and others. Files are self-describing and embed WebAssembly decoder binaries so a reader can decode data even without a matching native decoder, with FlatBuffer-defined structure and user-defined Wasm encodings for extensibility. It is MIT-licensed and described as a research prototype, not production-ready.
  • Comments: HN commenters asked for a concrete "why" over Parquet and ORC beyond the abstract claims, noted limited tool support, and flagged the repository had no commits in roughly eight months.
  • Why it matters: Embedding decoders in the file targets the format-versioning and forward-compatibility problems that force rewrites across Parquet and ORC deployments.
  • Follow-up: Track whether the prototype gains active development, benchmarks against Parquet/Lance/Nimble, and any engine adoption.

Send feedback on this story