Copies & Clones

Two files pass every identity test. Then someone edits one.

Duplication

Read the bytes. Write them somewhere else. Done.

A copy nobody touches is not a derivative of anything. It’s clutter.

Derivation

The claim that a copy now has its own purpose, its own audience, its own errors.

Nothing in the filesystem marks the transition.

Five people, five thresholds

  • one edit
  • a meaningful edit
  • independent circulation
  • a different audience
  • enough divergence that reconciling would take real effort

None of these is a property of the file. Two can be satisfied while the bytes stay identical.

Why there’s no test for it

Derivation is a claim about how a file will be used from now on.

The bytes record what has happened. No property of what has happened can encode what is about to.

The judgement has consequences a hash can’t settle: who owns the result, which version a reader should trust.

Librarians split the difference

Cataloguing gives one creation four separate levels:

  • work: the creation itself
  • expression: one realisation of it
  • manifestation: one published form of that
  • item: the single copy on the shelf

Four words because one word could not carry it. The filesystem has one.

The cataloguing model's four levels (work, expression, manifestation, item) each realised in the next, beside the filesystem's single level: a file.

Next week

Two related, unreconciled versions. What people do about it.

Usually without ever using the word “branch.”

Sources

  • Functional Requirements for Bibliographic Records: Final Report IFLA Study Group on the Functional Requirements for Bibliographic Records, 1998. The work / expression / manifestation / item model: a discipline that had to formalise when two things are the same creation, and needed four levels to do it. ifla.org