Formats & normalisation
- View
File format is the single biggest determinant of whether a digital record will still open decades from now. Some formats are open, well-documented and widely supported; others are proprietary, undocumented and tied to one vendor's software. Preservation practice manages this risk through format identification, characterisation and, where justified, normalisation to sustainable formats.
The first step is knowing exactly what you hold. Tools such as DROID, Siegfried and FIDO identify formats against the PRONOM registry, returning a persistent identifier (a PUID) rather than trusting the file extension, which is easily wrong or spoofed.
Choosing preservation formats
- Prefer open, standardised, non-proprietary formats with published specifications and multiple implementations.
- Common preservation targets: PDF/A (ISO 19005) for documents, TIFF or JPEG 2000 for images, WAV/BWF for audio, FFV1 in Matroska for video, CSV or well-formed XML/SIARD for data and databases.
- Weigh openness, adoption, complexity, transparency and the risk of embedded dependencies (fonts, codecs, external links).
Normalisation
Normalisation converts an ingested file into a preservation format at the point of ingest. It reduces the number of format "families" you must sustain and defends against obsolescence, but it is a transformation - so it must be done carefully.
- Always keep the original alongside the normalised copy; normalisation can lose significant properties.
- Record the migration as a PREMIS event: the tool, version, settings and the outcome (success/failure).
- Validate the output (e.g. with veraPDF for PDF/A, JHOVE for format-level validation) - never assume the conversion was faithful.
Key takeaways
- Identify formats by signature against PRONOM (DROID/Siegfried/FIDO), never by file extension.
- Favour open, documented formats - PDF/A, TIFF, JPEG 2000, WAV/BWF, FFV1/MKV, CSV/SIARD.
- Normalise to preservation formats but always keep the original, validate the output, and log the PREMIS event.