XeLaTeX, hyperref and PDF/A

Question

I am finishing my doctoral dissertation using XeLaTeX and facing the requirement to submit a PDF/A compliant document.

First, I want to emphasize that I have found and read several related posts (e.g. this and this). The solution based on using pdfx package seems not to be working with XeLaTeX and produces a series of errors starting with "Undefined control sequence. \pdfminorversion". So, I have pretty much followed this tutorial and succeed to generate a PDF/A compliant document using Ghostscript.

However, I can do this only when not using hyperref package or when it is called with the option draft. Like that, I obtain a PDF/A document without nice stuff: bookmarks, links and the option to be opened with 100 % zoom. When I try doing the same thing with hyperref package and option pdfa, I get this error when running Ghostscript: "Annotation set to non-printing, not permitted in PDF/A, reverting to normal PDF output". The links and bookmarks are kept, but 100 % zoom is not. (This is similar as when e.g. Word 2007 builds a PDF/A document.) But the file I obtain is not a PDF/A file.

This is how I call hyperref package from the preamble:

\usepackage[%draft=true,
            pdfa=true,
            bookmarks=true,bookmarksopen=true,bookmarksopenlevel=1,
            unicode=true,
            pdfauthor={My Name},
            pdftitle={My Title},
            breaklinks,hidelinks,
            colorlinks=true,linkcolor=blue,citecolor=blue,urlcolor=blue]{hyperref}
    \hypersetup{pdfstartview={XYZ null null 1.00}}

My question is how to keep the links and bookmarks, and if possible 100 % zoom, and get PDF/A file. Is it possible to set the options in hyperref package to accomplish something like that using XeLaTeX and, eventually, Ghostscript?

Answer 1

The latest version (1.5.8) of the pdfx package has been used to create valid PDF/A documents using XeLaTeX. However there are some difficulties.

Firstly, you must use an invocation of:

xelatex -shell-escape -output-driver="xdvipdfmx -z 0" <filename>

which means that the resulting PDF uses no compression, so can be more than 10x the size that it would be otherwise, if compression were used. This is because PDF/A does not allow the Metadata packet to be compressed, but there is currently no way to apply compression (or non-compression) selectively. (The -shell-escape is to allow use of texlua to generate some required information.)

Secondly, particularly when using CJK fonts, there seems to be no way to generate a valid /CIDToGIDMap. If this assertion is incorrect, then please let me know how, so I can include appropriate procedures in the next version of pdfx.sty. There is the possibility of setting /CIDToGIDMap /Identity but how does one determine whether this is appropriate?