The cdb package may be used to create and manipulate cdb's: "constant databases".

A cdb is a persistent-store file format developed by Daniel J. Bernstein, designed for fast access to keyed data. It is commonly known as a "hash" database --and also as an "associative array"-- where the physical location of a particular record may be directly computed from the value of the key.

cdb's are most suitable for applications where the data remains relatively constant, keys are known, and data access is frequent and generally read-only. That is, things like a user/password database, a chart of accounts for a general ledger, or a periodic table of the elements. On the other hand, you generally would not use a cdb to record minute-to-minute changes in meteorological data.

We've seen many .cdb's along the djb way, including:

• tcprules access rule definitions for services run with tcpserver
• the root/data.cdb data file for tinydns
• mail delivery aliases in /etc/aliases.cdb for fastforward

A qmail installation may include even more:

• the data file for qmail-users(9) in /var/qmail/users/cdb
• supplementary rcpthosts for qmail-smtpd in /var/qmail/control/morercpthosts.cdb
• message rewriting rules for ofmipd (in the mess822 package)

cdb's are used in these instances because they are fast and reliable. The cdb package described here let's us easily develop and use cdb's in our own applications.