Hi all, Just a quick update on the PR: https://github.com/apache/parquet-format/pull/585
One of the contention points on the PR was around how to reason around compression for data that's inside the Parquet file. The existing wording was: > The bytes referenced by a self-reference use the `CompressionCodec` defined by the `inline` column chunk's `ColumnMetadata`. Since a lot of folks in the community left comments around this, I wanted to update the VOTE thread to make interested parties aware of the change. We've removed this line. Essentially, we won't perform any additional compression for data that's self-referenced in the file. The rationale for this change is as follows: 1. Most multi-modal data is already compressed; jpeg, png, video formats, audio formats, pdfs, these formats already perform some form of compression. Double compressing can be a waste of cpus 2. We don't want to re-encode files when they're moved from file to file 3. We don't want to have a difference between external and self-referenced representations with regards to compression It seems like other formats also don't perform any additional compression on their blobs. Please let me know if you have any concerns around this change. Otherwise, I strongly believe that the change is ready for merging. I've addressed all remaining comments. Thanks, Burak On Fri, Jul 17, 2026 at 5:42 PM Burak Yavuz <[email protected]> wrote: > Hi everyone, > > The vote passes with 18 +1s (4 binding)! > Thank you to everyone who provided feedback and helped us land here. > > I will continue to wait for feedback on the PR in parquet-format for a few > more days. > Talat and Gaurav, let's please continue to chat on the design doc for > version identifier. I'll also wait until I get approval from Rok, Antoine, > and Gang, as they had left comments on the PR before. > > (* = binding) > +1: > Daniel Weeks (*) > Russell Spitzer > Eduard Tudenhöfner > Divjot Arora > Anurag Mantipragada > Burak Yavuz > Steve Loughran > Xuwei Fu > Andrew Lamb (*) > Ed Seidl > Talat Uyarer > Gunnar Morling > Prateek Gaur > Gang Wu (*) > Fokko Driesprong (*) > Kevin Liu > Neelesh Salian > Alkis Evlogimenos > > 0: None > > -1: None > > Best regards, > Burak Yavuz >
