Preservation Metadata

Metadata is what turns a pile of files into a recoverable collection. Use YAML for human-editable records and SHA-256 or stronger for content verification.

Artifact record

schema_version: 1
id: ffxiv-global-en-2012.09.19.0001-2026-07-16
kind: client                 # client, server, source, tool, document, database, patch
title: Final Fantasy XIV 1.23b client
version: 2012.09.19.0001
platform: windows-x86
region: global
languages: [en]
acquired_at: 2026-07-16T00:00:00+02:00
source:
  type: physical-media      # url, git, physical-media, personal-backup, unknown
  locator: restricted       # Never place secrets in this field
  acquired_by: maintainer-id
rights:
  access: restricted        # public, restricted, private, quarantine
  redistribution: not-authorized
  license: NOASSERTION
integrity:
  algorithm: sha256
  manifest: SHA256SUMS
relationships:
  - type: compatible-with
    artifact: project-meteor-<commit>
  - type: derived-by
    artifact: patchset-id
notes: >-
  Factual context only. Put uncertain statements in a linked finding.

Provenance requirements

Every artifact must answer:

• What is it?
• Where did it come from?
• When and by whom was it acquired?
• Are these the original bytes or a derived form?
• How can its integrity be checked?
• Who is allowed to access or redistribute it?
• Which server, client, tool, patch, or document does it relate to?
• Has it ever been successfully restored or run?

If an answer is unknown, use unknown; do not guess. Open a finding to investigate important gaps.

Checksum practices

• Hash original containers before extraction.
• Produce a deterministic recursive file manifest for extracted trees.
• Sort manifest paths consistently and document path encoding/case behavior.
• Keep checksum manifests with metadata and copy them to a separate catalog repository.
• Recheck after transfer, periodically on storage, and during restore drills.
• A matching checksum proves byte identity, not authenticity. Provenance is still required.

Time and identity

Use ISO 8601 timestamps with a timezone. Use stable maintainer IDs instead of email addresses where the catalog may be public. Preserve original filesystem timestamps as metadata when they are meaningful, but do not treat them as reliable acquisition dates.

Findings versus metadata

Artifact metadata contains established identity and provenance facts. Reverse-engineering observations belong in the Findings Log, where confidence and evidence can change without rewriting the acquisition record.