When using load balancing or clustering, it is important that requests do not rely on modifications made by previous requests. For example, a request could run a script that modifies a set command option. The set command option will be modified only on the Sous-processus Arbortext PE running on the Serveur Arbortext PE handling the request. A request to Arbortext Publishing Engine can't rely on any action performed by a previous request. Requests are fulfilled by multiple Sous-processus Arbortext PE that do not maintain a session state or record of a previous reques. Subsequent requests could go to a different Sous-processus Arbortext PE or a different Serveur Arbortext PE. This means that Arbortext PE Request Manager has no information to determine how to route subsequent requests based on previous ones.