Résumé
Exception handling and replication are two mechanisms that increase software reliability. Exception handling helps programmers control situations in which the normal flow of a program execution cannot continue. Replication handles system failures. Exceptions handling and replication do not apply in the same way to the same situations and thus are two complementary mechanisms to increase software reliability. The paper proposes a specification of an execution history oriented exception handling system for an agent language and middleware providing replication. This paper proposes an original signaling algorithm adapted to replicated agents and a rationale of how exception handling and replication mechanisms can combine to increase programmers' capability to achieve reliable agent-based applications.