J3/00-278
To:         J3
From:       /Interop
Date:       19 Sept 2000
Subject:    Issues 270 and 271
References: J3/00-235

Interop examined paper 00-235 and concluded that there were problems
with both proposed edits.

The suggested edit at 57:19, changing <declaration-type-spec> to
<type-spec>, neglects the constraints on the definition of
<declaration-type-spec>.  In particular, this change would have
removed the requirement that type-param-value expressions be
specification expressions.  This problem could be fixed
by duplicating the necessary constraint, but subgroup thinks
it "cleaner" to leave the reference as <declaration-type-spec>
and add a constraint against the CLASS keyword.

The suggested edit at 41:21 is incorrect and does not achieve its
stated goal anyway.  It is incorrect because it fails to account for
USE with rename.  There is no intent to disallow the use of a
<local-name> as in R1110.  It fails to achieve its stated goal anyway
because the bnf definition for type-name is simply that it be a name
(see the assumed rule in 1.6.3); there is no constraint that it be the
name of a derived type.  Note that R447 explicitly uses such a
constraint when it refers to <type-name>.  In fact, the wording in
R447 at 54:38 is almost identical to that at 41:21.  If it's a problem
at 41:21, then it's probably also a problem in R447.

On reconsideration, subgroup thinks it best to allow a type alias
for the parent type instead of clarifying that it isn't alowed.

[41:21] "the name" -> "a name"

[41:21] After "(5.5.3)" add "or of an accessible type alias for an
        extensible type"

[57:21+] Add new constraint

   "A <declaration-type-spec> in a <type-alias> shall not use the CLASS
    keyword."

[57:24-33] Delete issues 270 and 271.