SQLite compressed text-history prototypes

Simon Willison proposes an efficient method for storing text revision histories in SQLite by compressing a JSON array of all prior document versions using Zlib or Zstandard.
Initial prototypes demonstrated significant compression, reducing 20.4 MB of raw revision text (1,000 revisions) to just 80.3 KB when using Zstandard-compressed JSON.
To optimize performance and avoid decompressing the entire history on every edit, the scheme can be refined by breaking the compressed history into multiple SQLite rows, each holding a maximum of 128 revisions or 3MB of uncompressed JSON.