J3/04-277r1

Date:    4 May 2004
To:      J3
From:    Richard Maine
Subject: Sequence type fixes

The constraints for sequence types arguably allow
unlimited polymorphic, which was not the intent.
Proposed edit:

  [46:13-14 (C532)] "all ...sequence types" ->
     "each data component shall be declared to be of an intrinsic
      type or of a sequence derived type."

The edits proposed below are superceded by 04-289r1, which
has an edit along similar lines, but with improvements.

Although numeric and character sequence types can't have
type parameters, they accidentally allow components with
type parameters, which defeats the whole purpose of the
restriction.  The most robust way to fix this and also
avoid similar oversights is probably to define them
recursively instead of referring to ultimate components.
getting rid of the reference to ultimate components
also simplifies the wording because we no longer need to
circumlocute through words about objects of the type.
Proposed edits:

  [46:18] "and all ...of the type are" ->
    ", no pointer or allocatable components, and each
     component is"
  [46:19] "or default logical ... allocatable" ->
     "default logical, or a numeric sequence type"
  [46:20-21] "and all ...of the type are" ->
    ", no pointer or allocatable components, and each
     component is"
  [46:21] "and are not... allocatables" ->
     " or a character sequence type"