Checking the claim, before and after

Claiming a standard is one line inside the file. That line is not enough on its own: for the claim to be worth anything, an independent checker — one you run yourself, and one the team did not write — has to accept it.

Summary

The claim is checked at three moments.

  1. Before the file is written: you ask whether what you have composed would be taken, and get back a list of everything that stands in the way, all at once.
  2. While the PDF is being created: a document meant to claim a standard is refused unless it holds to everything that standard requires. Those checks are ours, and they run before a byte is written.
  3. Once the PDF has been created: veraPDF and Ghostscript — two free checkers that have never heard of us — open it and answer pass or fail.

The third one is the one that counts: only a checker from outside pronounces conformance. veraPDF, published by the Open Preservation Foundation and the PDF Association, two bodies whose business is keeping documents readable and in which the team owns no part, opens the file and accepts or declines it.

Technically

The report is asked for on the document as composed. Nothing is written and nothing is changed, and the document need not have claimed anything: the question is whether it would be taken. Each finding names the published text, what stands in the way, and what it is about — which font, which layer, which page, which link, which field, which cell.

The refusal at write time runs the same rules and stops at the first rule broken; it only ever touches a document that actually claimed a standard. Rules that ask for something missing — a tree at level A, a carried file at the level for carrying files — are refused separately and say what is missing rather than what is in the way.

A file the engine did not write can be read and reported against the same clauses. That report also lists what it did not read, and says why: what a link, a field or a page does when it is used is not followed, and the description packet is handed back as bytes rather than interpreted. Reading a document walks the resources of its pages, following nested drawings sixteen deep and stopping after 4 096 resource dictionaries.

Clause by clause

An empty report is not a certificate. What the engine does not know to look for, it does not report, and it never calls a document conforming: conformance is what a checker that did not write the file says.

What it does not look for, stated plainly so nobody counts on it: a password on a claimed document; transparency and blend modes; the contents of a carried file; the description packet itself; what the document does when something happens to it, as against what a page does; the marking of decorative elements; and a character the chosen font has no shape for. Each of those is a checker's to catch.

One difference between the two readings is worth knowing. Writing judges the document's font table, so a font declared and never used refuses the file; reading judges the fonts the pages actually name, so a font nothing selects is passed over.

Which standard a document declares is one of four editions. PDF/A-2 and PDF/A-3 rest on the same version of the PDF format; PDF/A-1 and PDF/A-4 each rest on another. Each has a page of its own:

What the engine does on its own

The engine does all of this as soon as a document claims the standard, whatever else the calling program asks for.

  • Reports everything it knows to look for at once, before anything is written, naming the published text and the subject of each finding.
  • Refuses to write a document that claimed a standard and breaks one of the rules above, naming the first.
  • Reads a file it did not write and reports against the same clauses, listing what it left unread and why.

What you supply

The engine states what it was given and invents nothing else: without those values it refuses the document rather than choosing one for you.

  • Asking for the report: nothing sets it off for you.
  • The checker's verdict, on the file that will actually be kept, which is the only verdict an archive accepts.
  • Everything on the list above that nothing here measures.

The other families of requirements

Which part of the archiving standard to claim Every font the document uses is embedded in the file Colours that mean the same thing on every device A page that comes out the same for everyone The document says what its marks stand for A document that carries other files inside it What the file says about itself

Back to what the archiving standard asks for

See the prices See the examples