[temp.arg.general] p2 template argument list of template-head seems to have a more restriction

The template argument list of a template-head is a template argument list in which the n_th template argument has the value of the n_th template parameter of the template-head. If the n_th template parameter is a template parameter pack ([temp.variadic]), the n_th template argument is a pack expansion whose pattern is the name of the template parameter pack.

Consider thie example

template<int I>
struct A{
    struct B;
};

template<int I>
struct A<I+0>::B{};  

The definition for class member B is rejected by most implementations. In this case, the nested-name-specifier is governed by [expr.prim.id.qual] p3

If A is the template argument list ([temp.arg]) of the corresponding template-head H ([temp.mem]), N nominates the primary template of T; H shall be equivalent to the template-head of T ([temp.over.link]).

Since I+0 can arguably be the value of the template parameter I, it should be the template argument list of the template-head. I don't know does it say has the value of the n_th template parameter intends to give the possibility to the above example? However, for the parameter pack, it uses the name. From the practice, we should also use the name for the common template parameter too. It may be that

The template argument list of a template-head is a template argument list in which the n_th template argument is the name(i.e. identifier) of the n_th template parameter of the template-head.

Start with the wording in [temp.arg.general] and compare it with [expr.prim.id.qual] p3, [temp.mem], and [temp.over.link]. Check how the A<I+0>::B example is treated by the draft and implementations. Done means the intended relationship between template parameters and template arguments is unambiguous in the standard text.