File formats decide whether your files outlive the software

This looks at file formats from the practical end — what holds up once conditions stop being ideal.

Documentation is what makes a format survive

A format with a published, complete specification can be implemented by anybody, which means readers continue to exist after the original software does not. Undocumented proprietary formats depend entirely on the vendor continuing to support them. This is why archives insist on open formats: not ideology, but the practical question of who can read the file in 2050.

The 1990s produced an enormous quantity of documents in formats that are now genuinely difficult to open.

Lossy compression is not reversible

JPEG, MP3 and most video codecs discard information permanently to achieve their size reduction. Re-encoding a lossy file compounds the loss, which is why a photograph that has been saved repeatedly looks progressively worse.

Keeping an original in a lossless format and exporting lossy copies is the practice that avoids this, and it costs storage. Editors that re-encode on every save compound it invisibly, so an image cropped one day and rotated the next is materially worse than the same two edits applied once to the original.

Containers and codecs are different

An MP4 file is a container that can hold video encoded in any of several codecs, plus audio encoded in others. This is why two files with the same extension can behave completely differently, and why a player may open one and refuse another. Diagnosing playback problems means identifying the codec rather than the extension.

Subtitle tracks, chapter markers and alternate audio live in the container as well, and conversion tools routinely discard them without saying so because they are copying the streams they were asked about.

Metadata is where meaning lives

Timestamps, camera settings, authorship and edit history are frequently the most valuable part of a file for future readers. Many conversion and upload processes strip metadata silently, which is sometimes a privacy benefit and always an information loss. Knowing which of your tools strip it is worth ten minutes of checking.

Filenames and folder structure are metadata too, and they are the layer that survives nearly every conversion, which is an argument for putting dates and subjects into them rather than relying on fields inside the file.

A workable personal policy

Text in plain text or a documented markup, images in an original lossless capture plus lossy derivatives, documents in PDF/A for anything that must be readable indefinitely. Avoid formats tied to a single application for anything you intend to keep. Migrate deliberately every few years rather than discovering the problem when you need the file.

Under load, recording checksums alongside the files is what separates a backup from a preservation copy, because storage decays silently and corruption is otherwise discovered only at the moment you need the file.

Implementations differ, and vendors are not obliged to document the differences.

The reader is the fragile part, not the file

A format survives only as long as working implementations do, and implementations rest on operating systems, libraries and hardware that all move underneath them. This is why material from obsolete systems is often recoverable only through emulation, in which an old environment is reconstructed well enough to run the original software. Emulation works well for widely used historical platforms and badly for anything niche, and it fails outright where the software checked a licence server that no longer exists.

Anything that must contact a company to authorise itself should be treated as unreadable in the long run whatever its file format promises.

The takeaway

Save it in something anyone can implement, and keep the original.

Understanding the failure mode tells you more than the feature list does.