If you are working with a header-only library, your mentioned approach will probably work. At least I can't think of any issue right away.
But if you have a compiled library that you have to link to, there is no way to put the library functions themselves into their own namespace (at least not without recompiling your own version of said library). That's because in the .dll or .so or what have you, each function has a mangled name that includes all namespaces (example). When you eventually link against the library, you can only "reach" those functions under that exact mangled name, which requires that your function calls are against that same (or, in your case, no) namespace as the compiled versions.
The "classic" workaround is to write a thin wrapper around the library, where for every exposed function, you do:
wrapper.h:
namespace libraryWrapper
{
void bar(int);
}
wrapper.cpp
#include "realLibrary.h" // Defines bar(int)
void libraryWrapper::bar(int x)
{
::bar(x)
}
Basic example