Exporting is where the migration becomes tangible. The goal is a complete, faithful, reproducible extract of the AtoM instance in formats the transform stage can consume. Faithful means the export preserves hierarchy, authority links and digital-object references; reproducible means you can run it again after a data fix and get the same shape of output.

AtoM offers several export routes, and the right choice depends on what you are moving and how much structure you need to keep.

Choose the right export channel

  • EAD 2002 XML - the richest route for archival hierarchy; one file per top-level description preserves the full nested structure and linked authority references.
  • CSV - flat, spreadsheet-friendly for descriptions, authority records and events; ideal for bulk field remediation but loses deep nesting unless you use AtoM's hierarchical CSV with legacy/parent IDs.
  • EAC-CPF for authority records and SKOS/CSV for taxonomies and place/subject terms.
  • OAI-PMH (Dublin Core or EAD) for harvest-style extraction where direct file access is limited.

Preserve identity and linkage

The most fragile things in any export are the identifiers that hold records together. Export legacyId, slug and parent/child references explicitly so the transform stage can rebuild the hierarchy and re-link actors. Never let AtoM re-mint IDs mid-export; capture the source primary keys as legacy identifiers you can carry through to Heratio and use for validation.

In practice. Snapshot the AtoM database and the digital-object store together, at the same moment, and freeze new cataloguing for the export window. An export taken while editors are still working produces a description set and a file set that no longer agree - the classic source of "missing image" errors after import.

Extract digital objects separately

Metadata and binaries travel on different paths. Export descriptions as XML/CSV, but copy the master files and derivatives from the AtoM uploads directory as a filesystem operation, recording each object's checksum. You will re-associate files to records by identifier during import, so the manifest linking filename to record must come out of this step.

Verify the export

Validate EAD against its DTD/schema, confirm record counts match the audit inventory, and check character encoding is UTF-8 end to end. A count that is even one record short means the transform stage starts from a broken baseline.

Key takeaways

  • Pick the export format by how much structure you must preserve - EAD for hierarchy, CSV for bulk field work.
  • Carry source identifiers through as legacy IDs so hierarchy and actor links can be rebuilt and validated.
  • Freeze editing, snapshot metadata and files together, and reconcile export counts against the audit.

Check your understanding

Last modified: Saturday, 1 August 2026, 11:10 AM