In a federated database system, a view mechanism is crucial since it is used to define exportable subsets of data ; to perform a virtual restructuring d ataset; and to construct the integrated schema. The view service in federated databa se systems must be capable of retaining as much semantic information as possible. The object-oriented ( 0 - 0 ) model was considered the suitable canonical data model since it meets the original criteria for canonical model selection. However, with the emergence of stronger object-relational (0 -R ) model, the re is a clear argument for using an 0 - R canonical model in the federation. Hence, research should now focus on th e development of semantically powerful view mechanism for th e newer model. Meanwhile, the availability of real 0 -R technologies offers researchers the opportunity to develop different forms of view mechanisms. The concept of roles has been widely studied in 0 - 0 modelling and development. The role model represents some characteristics that the traditional 0-0 model lacked, such as object migration, multiple occurrences and context-dependent access. While many forms of 0-0 views were designed for the 0-0 canonical model, one option was to extend the 0-0 model to incorporate a role model. In a role model, the real entity is modelled in the form of a role rather than an object. An object represents the permanent properties of an entity is a root object; and an object represents the temporary properties of an entity is a role object. The contribution of this research is to design a view system that employees the concept of roles for the 0 -R canonical model in a federated database system. In this thesis, an examination of the current 0 -R metamodel is provided first in order to provide an environment for recognising the roleview metadata and measuring the view performance; then a Roleview Definition Language (RDL) is introduced, along with the semantics for defining virtual classes and generating virtua l extents; finally, a working prototype is provided to prove th e role-based view system is implementable and the syntax is semantically correct.