= What we need =
 * More spatially redundant copies with active checks for file-integrity using checksums
 * Permanent unique access-ids with delimiters (foo/bar.zoo)
 * Permanent HTTP access point (e.g. https://storage.statsbiblioteket.dk/)
 * Write once, read many
 * Read partial files
 * Fast access ~ harddrive-like access time

= What we would like =
 * Access to checksums for files (caches, needed for content and link verification)
 * Graduated bit-preservation requirements (no backup / backup / two copies with checks / super paranoia / ...)
 * Shell access (read-only)

= What we don't know =
 * Rights & authorisation - can we somehow pass on authorisatoion to read a file? Is the URL safe for distribution?