Support PDF/A-compliant output using pdfx package

[dragonstyle]

Discussed in #4424

^{Originally posted by fuhrmanator February 18, 2023}
My publishing requirements ask me to generate a PDF that is PDF/A (archive) compliant. I did some googling, and found a great explanation at https://www.mathstat.dal.ca/~selinger/pdfa/. However, I wasn't able to get the metadata into the PDF file according to the documented process. I realized it's because it's hard to know exactly what the .tex file is named when pdflatex is invoked in Quarto. In my case, it's a book project, so I think the file was index.tex, even though the keep-tex results in a file with the name of the title of my book. I suppose it's something else when you write an article, etc.

Anyway, I read the docs for the pdfx package and found that you can embed the metadata into the .tex source, with a vanilla file name that "always" works. I tried it using header-includes: and it works well (or seems to). I see the metadata in the PDF, and Acrobat says my PDF is indeed PDF/A when I open it. I haven't fully validated the result with a real validation tool. But it's an encouraging start. Here's what I put in my _quarto.yml for a book project.

format:
  pdf:
    header-includes: |
      \usepackage[a-1b]{pdfx}
      \begin{filecontents*}[overwrite]{\jobname.xmpdata}
      \Title{Baking through the ages}
      \Author{A. Baker\sep C. Kneader}
      \Language{en-GB}
      \Keywords{cookies\sep muffins\sep cakes}
      \Publisher{Baking International}
      \end{filecontents*}

The first line loads the pdfx package, and the following lines create (overwrite) the metadata file, which is magically created temporarily in the pdflatex pipeline (so you don't need to worry about the exact filename, and it stays with your document).

Of course, you must duplicate some of the XMP info that is defined in other Quarto fields, e.g., Title, Author, etc. Maybe one day Quarto can make this happen automagically (in partials), but for now I'm happy I don't have to do a separate conversion step.

[tarleb]

Relevant pandoc issue: jgm/pandoc#5409

[paytonej]

Some nuance regarding whether #4426 and #13782 are duplicates... Issue #4426 is about PDF/A (long-term archival format) whereas #13782 is about PDF/UA (universal accessibility). The latter is about pdf tagging, which is not necessarily a requirement of the former.

Issue #13782 requires supporting the LaTeX2e \DocumentMetadata{...} declaration, which enables the new pdf tagging features for PDF/UA. The metadata declaration must appear before \documentclass{...} (see https://www.latex-project.org/news/2022/03/21/latex-dev-2022-1/). Quarto yaml header-includes: ... inserts text after the document class.

The pdfx package metadata mentioned in #4426 is not a viable solution for #13782 because it can't address the pdf tagging issue for accessibility; however, \DocumentMetadata{...} support to resolve #13782 would simultaneously address PDF/A metadata inclusion for #4426. In other words, addressing #13782 should resolve #8706, but not the other way around!

As pointed out by @jonassmedegaard in Accessibility mode for LaTeX (pandoc), support for \DocumentMetadata{...} is the most forward compatible solution to the tagged pdf issue. Further, support for \DocumentMetadata{...}, e.g. through partials, is likely to be the simplest path forward for addressing the LaTeX-based pdf accessibility issues in Epic: accessibility #8706, including #13248.