Floor 4
Conservation & Provenance
How sources are admitted, graded, and carried forward. The pipeline favours structured, citable corpora over broad but messy collections.
Provenance axes declared separately, never averaged
Structure
TEI XML and segmented JSON are preferred; plain text is accepted when work boundaries and metadata are stable.
Metadata
Every exported file receives frontmatter: author, title, language, sphere, source repository and path, counts, and export date.
Inclusion
Discovery is broad but inclusion is selective; the registry marks works for inclusion and defers unsuitable material.
Duplicate witnesses
When several repositories hold the same work, one canonical record is shown and the alternates are retained as provenance witnesses. The public interface shows the best current text first and exposes alternate sources as evidence — not as duplicate catalogue entries.