Delegation for Java: API or Language Extension?

The failed attempt of JavaSoft to incorporate an “Object Aggregation / Delegation API” into its newest JavaBeans model has demonstrated impressively the high necessity and also the notorious difficulty of incorporating delegation into typed class-based languages. Although JavaSoft's proposal has been withdrawn due to public criticism of its limitations, the general issue is still relevant: is it possible to define a “one size fits all” standard API for delegation, and if not, is there any real alternative to such an API? This paper explores both questions. It shows that on one hand, all API-level solutions have serious drawbacks related to functional limitations, simulation costs and sensitiveness to change. On the other hand, recent work has demonstrated that integration of dynamic delegation into a class-based, statically typed language with subtyping is feasible in theory and practice in spite of contradictory claims in literature. However, since an efficient implementation is still missing we can just recommend an API-level compromise and invite researchers worldwide to join efforts for a high-performance implementation of dynamic delegation.