To: J3                                                     J3/23-219r1
From: John Reid & Reinhold Bader & Malcolm Cohen
Subject: Interp on correspondence of unallocated coarrays
Date: 2023-October-25

/Interp declines to accept the submitted request in its current form.

The paper quite rightly says that the definition of coarray correspondence
means that unallocated coarrays do not correspond.

It then provides a problematic example with unallocated coarrays, and
claims that it conforms to the standard. But by the first point, that
unallocated coarrays do not correspond, it clearly does not conform to
the standard.

That is, the problem is the opposite of that stated: the problem is that
because unallocated coarrays are not established they do not correspond,
and therefore they can never be allocated via a dummy argument because
they cannot satisfy the correspondence requirement.

The focus of the paper on coarray components has also misled, because the
flaw is present for named coarrays. So the suggested approach for handling
the problem (add correspondence of non-coarrays) cannot work.

There are at least two other problems with the text in this area:
(1) there is an assumption that the "ultimate argument" is a named variable
    and not a subobject;
(2) the words "ultimate component" are used when it should almost certainly
    be "potential subobject component".

Work continues on exploring these issues and crafting an appropriate
response.

===END===