The bibliography is the download tree

Research-methods seminar, twelve students, one rule: every source gets fetched through the class curl service.

Each fetch is a job, and each saved file lands in storage reachable three ways at once: by-job/{uuid}, by-domain/{domain}, by-date/{YYYY-MM-DD}: symlinks over the same physical files. When the annotated bibliography is due, by-domain shows who read past the first search result, and by-date shows the student who fetched forty PDFs the night before.

Because the class's outbound HTTP is one service, its history is one tree. Citation checking stops being an honor system: the professor opens the storage listing. The provenance is the filesystem.

The bibliography isn't compiled at the deadline. It accumulates from the first search.