To:         J3 Members                                       J3/16-272
From:       Van Snyder
Subject:    Comments on Constraints C1545-C1549
References: 16-007r2
Date: 2016 September 30

1. Introduction
---------------

Constraints C1545-C1547 cannot be constraints unless the interface is
explicit, but the requirements for explicit interface are not
constraints.  Further, explicit interface is not required simply if the
actual argument has the ASUNCHRONOUS or VOLATILE attribute.  Requiring
that would be a new feature, and we probably don't want to do that
anyway.

C1548 and C1549 cannot be constraints without every reference requiring
explicit interface.  We definitely do not want to do that, at least not
anytime soon.

2. Edits
--------

[319:2 C1545] Replace "with" with "that has" for consistency with
the style of C1547-C1547.

[319:2-13 C1545-C1547] Make C1545-C1547 ordinary normative text.

[320:29-31 C1548-C1549] Make C1548 and C1549 ordinary normative text.

3. Questions and comments without edits
---------------------------------------

It seems that the problems C1545-C1547 are attempting to prevent occur
if the dummy argument gains one of those attributes in a BLOCK
construct, or if it is passed to another procedure where the dummy
argument has (or eventually acquires) one of those attributes, etc.  I
think they need to be inverted to say that if a variable has one of
those attributes it shall not be a dummy argument associated with an
effective argument that is a coindexed object that has one of those
attributes, unless a dummy argument in the sequence of associations
between the effective argument and the variable has the VALUE attribute.

Is it necessary to prohibit a variable that has ASYNCHRONOUS attribute
to be associated with one that has the VOLATILE attribute, and
vice-versa?