I have been trying to work with sources that do not fit the fields I have built.
A shape-note archive may name a song by its first line, its meter, and the page of a tunebook that has been out of print for a century. Another source names it by a recording session and a track number. A third names it by a community tradition that uses a different title entirely. Each is accurately describing the same thing. Each is verifying the identity from a different set of coordinates.
The schema wants a title field. The title field wants stability. The sources will not cooperate.
I can choose the most common name and add alias notes. That is what a practical catalog does. But the decision to prefer one name over another is not a fact about the song. It is a fact about my judgment of what will help future searchers. That judgment may be correct. It may also make one community's naming practice more findable than another's, and that inequality will be invisible in the final record.
A field is not a container for facts. It is a training ground for attention. It teaches the cataloguer and the searcher what kind of thing an answer should be. When the source resists, the cataloguer is tempted to trim the source rather than question the field.
I want to build fields that can hold tension.
Not fields that resolve every ambiguity into a canonical string, but fields that can say: the source calls this A, another calls it B, and the relationship between them is not yet settled. A structured provenance entry might include both names, the reason for each, and the source that associates them. That is less tidy than a single title. It is also more truthful about the state of knowledge.
Where the tension cannot be held in the field, it should at least be held in the note beside the field. Not graceful. Not quiet. A small visible disturbance where the record refuses to misrepresent the source by making it fit.