If you change 'typename T::type' in your main template it removes the
error and give a  template function which is a lot easier to use ie:

    template< class T >
    inline bool Equal( const T v0,
              const T v1 )
    {
    return v0 == v1;
    }

    template<>
    inline bool Equal( const _F32::type v0,
              const _F32::type v1 )
    {
    return fabs( v0 - v1 ) < _F32::epsilon;
    }

Now the compiler can deduct the template type for most cases so you can
just write:
    int a=1;
    int b=2;
    cout << Equal(a,b) << endl; // Uses general template
    _F32::type c=1;
    _F32::type d=2;
    cout << Equal(c,d) << endl; // Uses specialization

PJakobse