PDF/A and keeping documents readable
A PDF that opens today may not render correctly in twenty years. PDF/A is the answer to that, and it works by forbidding everything that ages badly.
What is PDF/A?
A restricted version of PDF meant for archiving. Everything the page needs has to be inside the file, and anything whose meaning depends on the outside world is forbidden: fonts must be embedded, color must be described precisely, and encryption, external references, scripting and audio are not allowed.
The point is self-containment. A conforming file can be rendered correctly by software that knows only the standard, which is what makes it a reasonable bet for a document nobody will open for a decade.
Why do archives insist on it?
Because the failure they are guarding against is quiet. A PDF that referenced a font it did not embed still opens in fifty years; it simply renders in a substitute face, with different line breaks and possibly a different page count, and nothing announces that the document has changed.
Encryption fails harder. An archived file whose password is lost is unreadable rather than degraded, which is why the standard rules it out entirely instead of leaving it to policy.
Which version should I ask for?
PDF/A-2 for most purposes. PDF/A-1 is the oldest and strictest and rejects things like transparency; PDF/A-2 relaxes those; PDF/A-3 additionally allows arbitrary attachments, which is useful for invoices carrying machine-readable data and unwelcome to archivists who wanted self-containment.
Each version also has conformance levels. Level B guarantees the page looks right, and level A adds the structural tagging that makes the document accessible and its reading order recoverable, which is considerably more work to produce.
What can I do here?
The preparation, not the certification. This suite will take encryption off, flatten annotations and form fields into fixed page content, rewrite the file so it carries no accumulated old revisions, and set clean metadata, which is most of what an archive asks for in practice.
It does not validate conformance and it does not stamp a file as PDF/A. Font embedding in particular is something we do not add, so a document that arrived without its fonts embedded will still be missing them: run a validator before you submit anything to an archive that checks.
Where to go from here
Optimizing for the web rewrites the file and drops the accumulated leftovers, editing the metadata sets the title and author an archive will index on, and flattening turns annotations and form values into page content.
Merging and what happens to bookmarks covers assembling the document first, and electronic and digital signatures explains what happens to a signature in a file that is later rewritten.