The Resurrected Registry: How to Read Between the Lines of a Deleted File

We talk a lot about archiving web pages, but sometimes the most crucial public record isn’t on a server; it’s on the hard drive of the computer across the desk. When a document is ‘deleted’ from a public office’s drive, or a draft version is overwritten by a final one, that’s often where the story ends. But there’s a quiet, technical form of digital preservation that happens in the margins, one that doesn’t require a web crawler, but a simple understanding of how word processors work. The technique? Forcing a corrupted file to confess its secrets.

The Unassuming .DOCX Autosave

Modern document formats like .DOCX or .ODT are not simple blocks of text. They are compressed archives, akin to ZIP files, containing XML files for the text, separate folders for images, and a critical component often overlooked: an autosave or recovery layer. When an application like Word crashes, it doesn’t just lose your work. It often saves a fragment of the document’s internal structure in a temporary file or within the document archive itself. When the user re-opens the file and clicks ‘save’ on the recovered version, that older data isn’t truly erased. It’s orphaned inside the new, clean file.

The practical trick is disarmingly simple. Take a suspect .DOCX file—perhaps one released through a records request that seems oddly brief or sanitized. Rename the file, changing its extension from .docx to .zip. Your operating system will warn you; confirm the change. Now, open that ZIP file. Inside, you’ll find a folder structure. Navigate to the ‘word’ folder. Alongside ‘document.xml’ (the main text), look for files with names like ‘_resaved’ or folders named ‘_old’ or ‘temp’. Sometimes, the treasure is in ‘/word/glossary/document.xml’ or buried in a ‘/_rels’ folder. These are the digital ghosts, the residual data from earlier edits that the software never fully purged.

This isn’t hacking; it’s reading the full artifact. You’re not bypassing security. You’re examining the document as it truly exists, not just as the word processor chooses to render it. I’ve used this to find paragraphs of explanatory text scrubbed from a final council memo, revealing the rationale behind a controversial line item. I’ve seen draft language with a staffer’s critical comments, saved as an embedded comment thread in the XML, that was absent from the PDF they released. The final, official version is a performance. The residual data is the rehearsal, and sometimes the rehearsal is more honest.

This technique forces a shift in perspective. Digital preservation isn’t always about grabbing a snapshot of a live website before it dies. Sometimes, it’s about being a more attentive reader of the files we already have, understanding that ‘deleted’ is a concept the user interface sells us, not one the file format always honors. It turns every document into a potential palimpsest, where the old writing faintly shows through. In an age of smooth, finalized PDFs, learning to read between these technical lines is a quiet act of archival rebellion, one that reclaims the whispers the machine was told to forget.

Notes & further reading

A few pages I came back to while writing this: