Saving is not enough to find again

A demonstration archive of six records distinguishes a file’s presence from the ability to recognise and retrieve it.

Digital anthropology
A drawer of envelopes and an index card linked to one envelope by blue thread.
A stored object needs a route for finding it again. AI-generated metaphor, no personal documents.

The file is there. We remember saving it, perhaps even the folder. When needed, though, none of the names looks right. A collection can be full and leave us almost empty-handed: it preserved the objects without preserving enough routes back to them.

For this page I built six fictional records for a small graphic project. Titles, dates and filenames are invented; no personal documents or metadata are exposed. We can search for a word and see which fields make it retrievable. This is not an archiving service and uploads nothing from the device.

A name carries only part of the work

A name like final-2 may suffice on the afternoon we choose it. Later its promise becomes ambiguous: final relative to which revision, prepared for which purpose? Adding another “definitive” preserves the history of haste, but not necessarily the history of the decision.

Six records to find again

Original comparison. The data and scenario were constructed for this page.

  • R01 · Poster draft

    2026-01-03 · First revision, not for delivery

    bozze/manifesto-01.pdf
  • R02 · Selected poster

    2026-01-05 · Paper printing, selected version

    stampa/manifesto-02.pdf
  • R03 · Poster for the web

    2026-01-05 · Digital publication

    web/manifesto-02.webp
  • R04 · Visual references

    2026-01-06 · Source origins and credits

    fonti/riferimenti.txt
  • R05 · Delivery inventory

    2026-01-07 · List of delivered files

    consegne/inventario.csv
  • R06 · Revision note

    2026-01-07 · Reason for the final choice

    note/revisione-02.md

Each record distinguishes identifier, title, date, purpose and path. Search considers these fields together without pretending to read a file’s contents. The word “print”, for instance, can retrieve a version intended for paper even when absent from its title. If a record lacks that information, the filter cannot invent it.

Finding again requires keeping some reason for recognition. The name helps; a brief description can add context a name alone cannot carry without becoming unreadable.

Describe while we still know

The Library of Congress resources on Personal Digital Archiving distinguish material types and include the issue of descriptions. This reference places personal archiving within a broader preservation task. Our comparison concerns only findability and context: it does not demonstrate a long-term preservation strategy.

Describing at the moment of saving may seem a cost imposed on an uncertain future. Yet that moment often holds information hard to reconstruct later: what an export was for, which copy should be delivered, why a version was rejected. A few relevant lines can suffice, provided they can travel with the material.

An image in a folder, a context card and a copy in a blue envelope.
Object, description and copy answer different questions. AI-generated illustration, not a technical backup diagram.

Tags can help create routes across folders. They can also accumulate synonyms nobody uses consistently. Before adding many, it helps to ask which questions will actually return: am I looking for an author, a date, a format, a right of use? A useful vocabulary grows from those questions, not the number of labels available.

A route that can break

The path says where to seek the file in the described system. If we change folders without updating the index, the record preserves a direction that no longer leads to the object. A stable identifier at least lets us discuss the same element when its address changes; it does not automatically repair a broken link.

Search has a boundary too. This page’s small archive offers no automatic image recognition, semantic similarity or spelling correction. It searches declared text. Making that limit visible is more useful than letting users believe that no results means no document exists.

Preservation is another check

A usable index is not a backup. The records could be perfect and the files unreadable; we might have several copies without knowing which contains the required version. Integrity, ability to open files and copies in distinct places require further checks this example does not perform.

The collection need not become a monument to order. It should leave someone returning a few reliable handholds. A saved file says something was retained. An archive begins when we can understand what it was and find a route to meet it again.

An image in a folder, a context card and a copy in a blue envelope.
Object, description and copy answer different questions. AI-generated illustration, not a technical backup diagram.