Several methods exist for updating HEPDB databases. Normally, updates are not applied directly to the database itself but queued to a dedicated server. The client-server communication also has several variants but the most important is when the communication is via files. Only this method will be described here as it is the only one enabled in the standard CERNLIB distribution of the package.
When a user accesses a HEPDB database a journal file is opened in which database modifications are written. This journal file is created in the user's directory and is in Zebra FZ alpha exchange format. This file is then moved to a queue directory upon request, when the user starts to modify a different database or when the database is closed. The queue directory is defined by a configuration file which is described in detail below.
The journal file name contains information on which database the modifications are for and the user and node name from which the update originated.
The above scheme works well in both localised and distributed environments and is designed with file systems such as NFS and DFS in mind. In the case of nodes which do not offer network file systems, journal files are transferred using the CERN Program Library package CSPACK [6].