The Library of Congress lists what the archival profile rules out. “All fonts must be embedded and also must be legally embeddable for unlimited, universal rendering.” “Audio and video content are forbidden. 3D artwork is also forbidden.” “Javascript and executable file launches are prohibited.” “PDF/A does not permit encryption.” The newest part relaxes some of this for engineering files: “PDF/A-4e supports Rich Media and 3D Annotations as well as embedded files.”

How is PDF/A different from a normal PDF?

A normal PDF may lean on things outside itself: a font it names but does not carry, a password, a script, a video. A PDF/A file gives those up, so it looks the same in any viewer years later, and it stays an ordinary PDF that every viewer opens. Each part is built on a PDF version; “PDF/A-1, the first PDF/A standard [ISO 19005-1:2005], was based on PDF version 1.4 (see PDF-1-4) and published in 2005.”

The other difference is a declaration. A file is PDF/A only when its XMP metadata says so: “This schema has two mandatory elements: pdfaid:part (integer) and pdfaid:conformance (closed list of text values).” A PDF that happens to follow every rule but makes no claim is not PDF/A, and a file that claims it can still break a rule.

What does the “2b” in PDF/A-2b mean?

The number is the part of ISO 19005 and the letter the conformance level: b for a file that looks the same everywhere, a for b plus tags that give the reading order, u for b plus a Unicode value for every character. Scans usually carry level b: “level B conformance is the level typically used for PDF/A files created from scanned pages.” PDF/A and editing has the table of every part and level.

Where do you see that a file is PDF/A?

  • In the metadata. PDF Textfix’s demo invoice embeds every font, yet it is not PDF/A: it has no metadata stream at all, so it claims nothing.
$ pdfinfo "Demo Invoice.pdf" | grep -E "Metadata Stream|PDF version"
Metadata Stream: no
PDF version:     1.3
  • In Acrobat, which opens a PDF/A file in a PDF/A viewing mode, read-only until you turn the mode off; PDF/A and editing shows how.
  • In a validator or preflight report, which checks the rules, not the claim.
  • In PDF Textfix’s banner, for example “PDF/A-2b document: characters missing from its fonts are always added with an embedded font.”

What does an edit do to a PDF/A file?

After any change, conformance has to be validated again, because the claim is only a statement the file makes. PDF Textfix keeps the identification, takes a missing character only from an embedded font, never from one the file merely names, and keeps a line break before every endstream. It does not validate the result; PDF/A and editing covers checking the saved copy.

Questions about PDF/A

What does PDF/A compliant mean?

That the file follows every rule of the part and level it claims, which a validator confirms.

Is PDF/A the same as a linearized PDF?

No. Linearization orders a file for page-at-a-time loading; PDF/A limits what it may contain. A file can be either, both or neither.