PubSalvagePubSalvage

Release notes

What changed in each version, and the limits we know about. Newest first.

1.0.0: early access

Release candidate. The release date will be added here when the first public build is published.

What it does

  • inspect: checks whether each file is a readable Publisher file, which generation made it and what is inside. Free, with no limit, and it never writes anything.
  • render: rebuilds each page as SVG with a browser index and the text, and rates every file green, amber or red with reasons. --pdf adds a PDF, --odg adds an editable drawing meant for LibreOffice Draw, and --report adds CSV, JSON and HTML reports for the batch. When a file's pages can't be rebuilt, or only some of them, its text and pictures are salvaged into its own folder.
  • rate: the rating and the batch report without writing pages.
  • salvage: takes the text and the pictures stored in the file (PNG, JPEG, BMP, WMF/EMF, PICT, TIFF) without the layout engine, with an optional manifest.
  • status, activate and licences: show the edition in use, store a licence key on your own computer, and list the licences of everything PubSalvage is built from.
  • --selftest-sandbox: proves on your machine that the layout-engine sandbox holds.
  • Old 8-bit text in Cyrillic, Greek, Hebrew, Arabic and Thai that other tools read as gibberish is detected and re-decoded; --codepage chooses the code page yourself.
  • Every file is opened in its own sandboxed process: an AppContainer and a Job Object on Windows, seccomp and Landlock on Linux.
  • The desktop app for Windows and Linux: add files or folders, choose where to save, press Rescue, read the results, and look at the rebuilt pages in a built-in viewer. It can scan a folder for Publisher files, shows the edition on a Licence page, works in light and dark, supports keyboard use and keeps no history.
  • Offline licence keys: Free, Personal, Pro and MSP editions, checked on your computer with no activation server.

Known limits

  • Ratings are not yet measured on real Publisher 2003-2021 files. We have tested on very few real files. Green means every word was found on the rebuilt pages and every other check passed (any lost word makes a file amber), not that the page looks identical to Publisher's. Look at the pages that matter to you.
  • The PDF output can't draw Hebrew, Arabic, Thai or CJK text (the text file and SVG pages keep it), flattens gradient, picture and hatch fills, and doesn't reproduce clipping or text effects.
  • The ODG export is new. It is checked structurally, but it hasn't been opened in LibreOffice on many real files, and links, equations and some styles aren't carried over. Treat the PDF as your reference copy.
  • Publisher 95/2.0 files aren't supported. Publisher 2002 to 2021 are reported as one family.
  • The text-overflow check is an estimate and can only add an amber reason.
  • Text order from salvage follows the text stream, not the page. Text from files made before Publisher 98 can include some font and printer names.
  • Files are processed one after another. Parallel batches aren't implemented yet.
  • The Windows sandbox's self-test has not been run on a Windows machine since the 1.0 changes, and it has not been tried on a range of customer machines. Windows installers aren't code-signed yet.
  • The free edition's limit is per run (3 files) and nothing is remembered between runs.
  • macOS isn't available. It is planned for version 1.1.

See Get PubSalvage for availability and how to verify a download, and the security overview for how the privacy promises are enforced.