Why do we have to use additional construction when iterating variadic template parameters?

Why do we have to use additional construction such as (non() - function, temp[] - array, or empty[](...){} - lambda) when iterating variadic template parameters?

As known, we can iterate a parameter pack with variadic templates in C++ by using some of these ways:

http://ift.tt/2aruJiM

#include <iostream>
#include <cstdlib>
#include <valarray>
#include <numeric>
using namespace std;

template<typename ...Args> constexpr inline void non(Args ...) {}

template<typename T, typename ...Args>
inline T sum1(T val, Args ...args) { non(val += args ...); return val; }   // v1
// why do we need some function non() here?

template<typename T, typename ...Args>
inline T sum2(T val, Args ...args) { auto tmp = { val += args... }; return val; }   // v2
// why do we need some array tmp[] here?

template<typename T, typename ...Args>
inline T sum3(T val, Args ...args) { [](...){}((val += args)... ); return val; }   // v3
// why do we need empty lambda [](...){} here?

template<typename T, typename ...Args>
inline T sum4(T val, Args ...args) { for(auto &i:{ args... }) val += i; return val; }//v4

template<typename ...Args, typename T = common_type_t<Args...>>
inline T sum5(Args ...args) { return std::valarray<T>({ args... }).sum(); }   // v5


template<typename T> constexpr inline T sum6(T val) { return val; }

template<typename T, typename ...Args>
constexpr inline T sum6(T val, Args ...args) { return val + sum6(args...); }  // v6


int main() {    
    cout << sum1(1, 2, 3) << endl;
    cout << sum2(1, 2, 3) << endl;
    cout << sum3(1, 2, 3) << endl;
    cout << sum4(1, 2, 3) << endl;
    cout << sum5(1, 2, 3) << endl;
    cout << sum6(1, 2, 3) << endl;

    return 0;
}

But why do we need to use:

1. non(val += args ...); instead of val += args...;
2. auto tmp = { val += args... }; instead of val += args...;
3. [](...){}((val += args)... ); instead of val += args...;

It would be much clearer and easier to use so:

template<typename T, typename ...Args>
inline T sum(T val, Args ...args) { val += args...; return val; }

Why is no there in the standard such possibility, or could such a possibility carry any danger?

And will be such possibility in C++17 or later?