News spool
The directory where a Usenet server kept every article — traditionally one file per article and one directory per group.
Under B News and C News the spool lived in /usr/spool/news, later /var/spool/news. The group comp.lang.c became the directory comp/lang/c, and each article a file named by a sequence number. It was transparent and easy to debug: with ls and grep you could read the news without a newsreader. A nightly expire job pruned whatever had grown too old.
The model fell apart as traffic grew. Millions of tiny files ran out of inodes long before the disk filled, and deleting them was a performance problem in its own right. INN 2.0 therefore introduced alternative storage methods, among them CNFS, which writes articles into large circular buffers that overwrite the oldest. Retention became a question of disk size rather than policy.