Almost every studio that was working before about 2005 has one of these. A magazine's complete run. A publisher's backlist. A council's newsletters. An agency's client work, kept because the contract said to. It sat on a server, then on a NAS, then on a stack of external drives in a cupboard, and at some point the question changed from "where is it?" to "can anyone still open it?"

The good news is that this is a well-understood kind of job. It has a shape, and the shape is: inventory, protect, triage, decide, produce, verify. Do it in that order and it is a week of work. Do it out of order and it is a year and a couple of losses.

The one rule everything else hangs off. Never let a conversion, a repair, or a "quick look" write to an original. Copy first, work on the copy, keep the source folder read-only. Every other mistake in this process is recoverable. That one is not.

1. Inventory — you cannot preserve what you cannot find

Get everything into one place and produce a list. Not a tidy place, not a sorted place — one place, with a list.

Search across every drive for the whole family of QuarkXPress extensions, not just .qxd: .qxt templates, .qxp projects from version 6 onwards, .qxb books, .qxl libraries, and .qwd and .qwt web documents. Then search again for files with no extension in the same folders, because Mac-authored work from the 1990s very often has none — see old Mac Quark files with no extension for why, and how to identify them.

While you are there, capture the two things people always forget and always need later:

  • The image folders. QuarkXPress kept a preview in the document and the real picture on disk. Documents without their images are half a record. They usually sit in a sibling folder, often called Links, Images or Pictures.
  • Any PDFs or printed proofs already in the archive. A PDF of how a page actually looked is often the most valuable object in the whole cupboard, and it is free — it already exists.

A plain spreadsheet is the right tool. Columns: path, filename, size, modified date, publication, year. You will add two more in a moment.

2. Protect the originals — properly, and verifiably

Copy the whole inventory to a Mac-format volume (APFS or HFS+) if any of it came off an old Mac, before you do anything else. FAT32 and exFAT volumes silently strip Mac file metadata, and every extra hop on one is another chance to lose it.

Then take a checksum of every source file, and keep the list:

find . -type f -exec shasum -a 256 {} \; > originals-sha256.txt

This takes minutes and earns its keep twice. It shows that nothing you did later altered a source. And rerunning it in five years tells you whether the storage itself has degraded — silent bit rot on old drives is real, and a checksum list is how you find out before the file is the only copy.

Set the source folder read-only. Then forget it exists and work from a second copy.

3. Triage by version — the column that turns a pile into a plan

Add a version column to the spreadsheet. This is the single most useful hour in the project, because what is possible for each file is entirely determined by the version that wrote it.

You can read it from the first bytes of each file without any special software, on Mac or Windows — the method is in which version of QuarkXPress made this file? The same check tells you whether each file was Mac-authored or Windows-authored, which matters for some conversion routes.

What you will typically end up with:

  • 3.1 – 4.1 — the best-served range. Free readers exist; several routes out.
  • 5 — the awkward one. Quark's converters or a commercial plug-in.
  • 6 and later — projects rather than documents; Quark's own path is the realistic one.
  • Not a QuarkXPress file at all — there are always some. Old PageMaker, stray previews, or something truncated in a 2003 copy.

Add a second column: size sanity. Anything implausibly small is a failed copy from years ago, and you want to know that now rather than after you have built a workflow around it.

4. Decide: preserve or rebuild

This is the decision that saves the most money, and most people never consciously make it.

Preserve means: keep the work readable and findable forever. The output is text, page structure, and a visual record. Nobody will edit these again. This is the right call for a complete magazine run, a council's newsletters, a closed backlist, anything kept for reference, rights or heritage — which is the overwhelming majority of a typical back catalogue.

Rebuild means: produce a live, editable modern layout. That is a designer's project per document, with fonts to substitute and pages to repaginate. It is worth it for a title actually being reissued. It is not worth it for issue 143 of a newsletter from 1998.

Be ruthless. A realistic split is a small handful of rebuilds and everything else preserved. If you try to rebuild the lot you will finish neither.

5. Produce the archive — what to actually keep

For each preserved document, aim for four things sitting side by side in one folder:

  1. The original file, untouched. Always. Formats become readable again; tools improve. What cannot be recovered is a source nobody kept.
  2. The text, as plain text. The most durable artefact you can make. No application, no licence, no version. Searchable in any decade.
  3. A visual record of the pages — PDF if you can produce one, page images if you cannot. This is what answers "what did it look like?" and it is what most people actually want from an archive.
  4. A structured layout file if you can get one — IDML, say — for the day someone does want to rework a page.

Plus one file per publication: a plain-text or CSV manifest listing every document, its version, what was recovered from it, and what was not. In five years the manifest is what makes the archive usable rather than a folder of mysteries.

Where the routes come in

For the 3.1 – 4.1 pile, free options genuinely cover a lot of ground: LibreOffice Draw's QuarkXPress import filter reads that range on both platforms, and strings in a terminal will pull raw text out of a single file in seconds. Both, with their limits, are set out in how to open a .qxd file without QuarkXPress. For version 5 and above, Quark's own free converters are the realistic path. Try the free routes on a sample before you spend anything.

Where they get uncomfortable is scale and repeatability. Opening four hundred documents by hand in a GUI, one at a time, exporting each, and remembering what happened to each one, is not really a workflow — and it produces no record of what was recovered.

That gap is what QuarkXPress Rescue is built for. For QuarkXPress 3.1 – 4.1 documents, Mac- or Windows-authored, it reads each file locally and read-only, recovers the page layout geometry and the authored text stories with the font, point size, bold/italic and colour of each run, pulls out the pictures stored inside the file it can positively identify as ordinary image files, and writes plain text, a styled RTF and a structurally-valid IDML package along with the referenced font and colour names — with a recovery report on every run stating what it found. That report is the per-file record a manifest is built from. Nothing leaves your machine and there is no account. The app runs on Windows 10 and 11 (64-bit) and on macOS 10.15 or later (Apple silicon and Intel Macs); one purchase covers both.

Its limits, so you can plan around them rather than discover them. It reads QuarkXPress 3.1–4.1 only. Version 5 files are identified but not decoded; versions 6 through 2019 are identified precisely by version and reported as not recoverable. Paragraph settings are decoded for the version 4 family but not for the version 3 family, and custom colour display names are not decoded. Picture extraction is proven on a Mac-authored (PICT) 4.1 document; the Windows BMP/WMF path is written but not demonstrated on a genuine Windows image-bearing file, and a payload that cannot be positively identified is listed as present but not extracted rather than guessed at. On the one multi-page Mac 4.1 document we hold, the pages, the ten text boxes and the embedded picture came out but no text story decoded — so sample your own multi-page files before planning a run around them. Exact layout and colour fidelity is confirmed by opening the IDML in InDesign, which is a human check at your end. For a mixed archive, expect it to handle one pile and Quark's own converters to handle the other.

6. Verify, then store like you mean it

Two checks before you call it done.

Spot-check the output against the record. Pull ten documents spread across the years, and compare the recovered text and pages against the PDF or the printed copy. You are looking for missing stories, not for perfect line breaks. Ten files is enough to catch a systematic problem.

Re-run the checksums on the source folder and confirm the list matches the one you took in step two. That is your evidence that the process was non-destructive — and if you are doing this for a client, an institution or a rights holder, it is the thing to put in the report.

Then store it properly: two copies on different media in different places, one of them offline, and a diary note to verify the checksums annually. Every large archive that has been lost was, at the time, on a drive somebody was fairly sure was fine.

The test to hold it to

When you think you have finished, ask one question: if the studio closed tomorrow and someone unfamiliar with any of this opened the archive folder in 2046, could they read the work?

Plain text, PDFs and a manifest pass that test. A folder of .qxd files and a note saying "ask Dave" does not.

This is genuinely worth doing before it becomes urgent. There is no deadline here and we are not going to invent one — just the ordinary day someone needs the words back, and finds that the last machine that could produce them was recycled in 2019.