Keeping documents readable for decades

PDF became the default archival format for a reason: it fixes appearance in a way editable formats do not. But a file that opens correctly in twenty years takes slightly more care than one that opens today.

Embed the fonts

A PDF can either embed the fonts it uses or reference them by name and hope the reader has them. The second option produces smaller files and a document that renders differently, or not at all, on a machine without those fonts.

Most modern export pipelines embed by default. It is worth confirming for anything being kept, because substituted fonts change line breaks, which changes pagination, which changes what page a clause appears on.

Prefer text over images

A scanned document is a picture. It cannot be searched, indexed, quoted, resized usefully or read aloud, and it degrades every time it is recompressed.

Where you have the choice, archive the digital original rather than a scan of a printout of it. Where you do not, add a text layer through recognition so the content is at least findable.

PDF/A exists for this

There is a subset of the specification, PDF/A, designed specifically for archiving. It requires embedded fonts, forbids external references, forbids encryption, and forbids anything that depends on software behaviour rather than the file's own contents.

The requirements are essentially a list of the things that break long-term readability. Even if you do not formally produce PDF/A, its rules are a good checklist.

Avoid anything that depends on something else

External links expire. Referenced content moves. Embedded video needs a codec. JavaScript needs an interpreter that still behaves the same way.

A document that will be read in decades should be self-contained. Anything it needs, it should carry.

Do not encrypt archives

Encryption and long-term storage are in direct tension. The password will be lost. Not might be, will be, on any horizon long enough to matter. If the archive needs protection, protect the storage rather than the individual files.

Keep the originals

Every conversion loses something. Compression discards image data, recognition guesses at characters, conversion to another format infers structure. None of it is reversible.

Where storage allows, keep the original alongside the processed version. Storage is cheap; the document you cannot reconstruct is not.

Check them occasionally

Files rot quietly. A backup nobody has opened in five years is a hypothesis, not a backup. Opening a sample once a year turns a silent failure into a recoverable one.

Why formats fail

Not usually dramatically. A format falls out of use, the software that read it stops being maintained, and one day the file opens with the layout wrong and nobody notices because nobody is looking.

PDF is unusually safe here, because the specification is public, widely implemented, and old versions remain readable. That is precisely why it became the archival default.

The failures that do occur are specific and avoidable: missing fonts, external dependencies, encryption, and content stored as images with no text.

Fonts, in more detail

A PDF either embeds the fonts it uses or names them and hopes. On a machine without those fonts, a viewer substitutes something else.

Substitution changes character widths, which changes line breaks, which changes pagination. A contract where clause 14 was on page 6 now has it on page 7, and every cross-reference in the correspondence around it is wrong.

This is the single most common way an archived document degrades without appearing to be damaged.

What PDF/A requires, and why

The archival subset forbids the things that break long-term readability: it requires embedded fonts, forbids external references, forbids encryption, and requires colour to be defined absolutely rather than by reference to a device.

Even if you never formally produce PDF/A files, that list is a good checklist. A self-contained, unencrypted, font-embedded document with a text layer will still open in thirty years.

Storage practices that matter more than format

Keep more than one copy, in more than one place. A single archive is a hypothesis.

Check them. A backup nobody has opened in five years may already have failed. Opening a sample annually turns a silent loss into a recoverable one.

Record what things are. A folder of files named after their original filenames is a puzzle a decade later. A short index costs an hour once.

Keep originals alongside processed versions. Every conversion loses something, and none of it is reversible.

Encryption and archives do not mix

On any horizon long enough to call archival, the password will be lost. Not might be. Will be, because the people who knew it will have moved on and the note recording it will have been in a system that was decommissioned.

Protect the storage rather than the individual files. An encrypted archive with a forgotten password is indistinguishable from a deleted one.