It is also clear that some definition of the behaviour of existing standard C++ libraries will be required. Many of the existing C++ functions will require definition of behaviour within parallel systems - and this behaviour may be different within different parallel architectures. As an example, consider the gettimeofday() function in a distributed system. This could, and indeed is likely to return different values in different nodes of a distributed system. A problem such as this could be left to applications programmers to handle, but it would be more useful for EC++ to define how these functions should behave in various parallel architectures. These definitions may or may not require a re implementation of the functions in a particular hardware implementation of EC++, but it would appear to be an issue which Europa should decide to address.