Fundamental to the operation of HEPDB is the concept of the database server. Although other modes of operation exist, the normal way of updating a HEPDB database is via a dedicated database server. The client, or user, requests changes to the database using the HEPDB Fortran or KUIP interface. These changes are not applied directly. Instead they are written into a queue, which might simply be a directory or the VM spool. Thus although the database files may be read by an unlimited number of simulatenous users, they are only ever written by a single process - the database server.