What is PDF/A?
PDF/A is ordinary PDF optimized for long-term archiving. It is defined by ISO 19005, and its single goal is that a document renders the same way in the far future as it does today.
The idea behind PDF/A
A PDF/A file must be self-contained and unambiguous. Every rule in the standard follows from that. If rendering the page requires something the file does not carry, or something whose interpretation depends on the machine doing the rendering, PDF/A rules it out.
That is why fonts must be embedded: "Helvetica" means one thing on the machine that produced the file and something subtly different, in metrics and in glyph coverage, on the one that opens it. It is why device colour needs an OutputIntent: "50% cyan" is not a colour until you say which press or screen it is for. And it is why encryption is forbidden: a document nobody can decrypt in a few years is not archived, it is lost.
What a conforming file must carry
A conforming file must carry every font it uses, embedded, with the glyphs actually used and enough encoding information to map codes to characters, an OutputIntent with an ICC destination profile, whenever device-dependent colour appears anywhere in the document, and XMP metadata identifying the PDF/A part and conformance level, consistent with the document information dictionary.
What a conforming file rules out
- Encryption of any kind, including an owner password with an empty user password.
- JavaScript, launch actions, and anything else that executes.
- External content: linked fonts, referenced streams, embedded video and audio.
- Transparency and blend modes, in PDF/A-1 only. PDF/A-2 allows them.
- LZW compression, and non-embedded or non-symbolic-safe font encodings.
Each of these has a corresponding entry in the error reference, with the clause it comes from and how to resolve it.
Conformance is checkable, correctness is not
PDF/A conformance is a mechanical property of the file. A validator can decide it, and two validators that agree with the standard will agree with each other, which is exactly why a corpus like Isartor can exist. What no validator can tell you is whether the document says the right thing, whether it is the version you meant to keep, or whether converting it to PDF/A changed how it looks. That last one is a real risk in conversion, which is why gopdfrab has a separate fidelity check.
Which version should I target?
There are four PDF/A parts (PDF/A-1, PDF/A-2, PDF/A-3, PDF/A-4) and several conformance levels. PDF/A-1b is the most widely accepted. PDF/A-3 is what e-invoicing is built on. Find out more about it here: parts and conformance levels.
Frequently asked
Is PDF/A a different file format from PDF?
No. A PDF/A file is an ordinary PDF that additionally satisfies the constraints of ISO 19005. Any PDF reader can open it; the constraints exist so that a reader decades from now can render it the same way as one today.
What does PDF/A forbid?
Anything whose meaning depends on the environment or on time. Fonts that are not embedded, colour that is only meaningful with an unknown output device, encryption, JavaScript and launch actions, external references, audio and video, and transparency in PDF/A-1 are all disallowed.
Does PDF/A guarantee my document looks the same forever?
It removes the main reasons a renderer would disagree with the one you used. It cannot guarantee anything about a viewer that ignores the spec, and it says nothing about whether the content is correct, only that it is reproducible.
Do I need PDF/A?
If a document has a retention obligation (invoices, contracts, court filings, records deposited with an archive) PDF/A is often required and the right choice.