What are the differences between UNO and CORBA?
UNO does not generate stub code. We did this in the former UNO2 implementation. One problem was, that the generated stub code was quite a heavy load (about 12 MB at former times, even though the API had less types as compared to now!). With UNO3 (the current version), we decided to generically call implementations, without having generated stub code, just by providing bridging code that knows the specific compiler that compiled the implementation. The only thing that is generated for C++, besides the pure virtual classes, are the getCppuType() functions to reflect types. The full type description is read at runtime from a binary storage. This approach has the advantage that on increasing numbers of types, less space is used. To see more the differences between UNO and CORBA, please have a look on the following comparison.