In this work, distributed and sharable learning resources are modeled by two types of Learning Objects (LOs): Atomic Learning Object and Composite Learning Object. LOs are uniformly published as Web-services in a constraint-based Web-service registry and are made sharable and reusable. This paper presents the learning object models for the specification of these two types of LOs and an extended Web- service infrastructure, which provides a standard framework for the registration, discovery, binding and invocation of these objects. An Event-Trigger-Rule Server is integrated with a Learning Process Execution Engine to make Composite Learning Objects active, flexible, customizable and adaptive.