Main idea

Faced with a file to put away, the reflex is to look at its subject. A translation guide talks about the English translation of a book, so it gets put next to the book, with the manuscript, the language folders and the scripts. The criterion used is the subject, and it looks so obvious that nobody even states it.

A second criterion exists, and it doesn't give the same answer: what does this document do? The translation guide is neither a chapter, nor a translation, nor an appendix; no reader of the book will ever read it. It doesn't add to the results of the work, it determines how the next results are to be produced — which criteria to apply when a sentence sounds too literal, where to stabilise a term that shifts from one chapter to the next, which checks to run before accepting a translation.

The two criteria part company exactly there: the guide doesn't describe what the work has produced, it changes the way that work is to produce. Filed next to the book, it becomes one more document sitting alongside the others; filed at the head of the workspace, it becomes a rule binding on every translation that follows.

Why it matters

The subject criterion and the function criterion coincide as long as you only handle anticipated objects: a source talks about the subject and serves as a source, a chapter talks about the subject and is a chapter. They diverge as soon as a document appears that talks about a piece of work without being part of it — an editorial charter, a naming convention, a review grid, an interview protocol. Filing by subject buries them in the middle of the material they govern, where nobody reads them at the right moment.

The question that settles it fits in one sentence: does this document produce a result, preserve knowledge, describe the current state, or govern future actions?

Nuances and limits

A document's function isn't always visible at birth. The translation guide started as a list of remarks jotted down during a rereading; nothing, at that point, set it apart from a working note. It was the use we decided to make of it — reading it before each new translation — that moved it from one side to the other.

A single document can also carry both functions: a translation lexicon holds settled choices, which govern, and ongoing hesitations, which commit to nothing. Classifying it as one block means treating the latter like the former.

Open questions

• What to do with a document whose function changes after the fact: move it, at the risk of breaking the paths that reference it, or leave it where it is and lose the criterion?