In standard projects, commands are processed one at a time so that the next command cannot begin until the previous one has finished. In principle, the state of the system is therefore always known. In Global, remote commands are processed in parallel and so the next command may be initiated before the previous one has finished. This mode of operation is called non-blocking and its advantage in Global is to prevent a slow long-transaction command from blocking the user. Its disadvantage is that the user needs to work in a new way to exploit this parallel nature of Global.

If a remote command traversing the Global network becomes held up at a particular location (for example due to a comms line fault) then, for most commands, the command is placed in a transaction database at that location for later processing. A small number of commands, known as ‘kernel’ commands, bypass the transaction database and are stored in a pending file for later processing. The use of the transaction database and the pending file means that commands are guaranteed to complete, but some commands may not succeed. Some may roll back, while others may just fail.

For further information about the transaction database, refer to Transaction Audit Trail, and Transaction Database Management.