The dataset is mounted into every student container once

The data science course uses a 40 GB corpus. Sixty students. The old plan was sixty downloads on the first day, a saturated campus link, and three students who ran out of disk.

The instructor's container holds the corpus at /data/corpus. One POST to /api/v1/containers/{id}/storage/shares with target_project_id set to the course project and mode: readonly. The share auto-mounts on every container in the project, now and for every container created later, at /hoody/shares/corpus/. Students open pandas against it; nobody can accidentally delete it.

When the instructor fixes a mislabeled column in week four, she fixes it once. Sixty notebooks see the fix on the next read.

Sixty students read one copy of the data, because the host mounted it sixty times.