(j3.2006) 14-007

[Van Snyder]

On Fri, 2014-01-24 at 12:07 +0900, Malcolm Cohen wrote:
> There are two new Unresolved Technical Issues, both arising from
> corrigendum 3. 
> Both contain suggested repairs.

I don't understand why the <selector> being a procedure pointer is a
problem, i.e., why was the answer to question (2)(a) in F03/0139 "no?".
Indeed, for those cases where the procedure pointer is a component of a
component... of a component, it would be useful to allow it, for the
same reason we want ASSOCIATE for data objects.  Rather than C805 (which
would simply be deleted), 8.1.3.3 would need some attention.  As is the
case with data objects, a procedure pointer would have to be associated.

If the <selector> can be a procedure pointer, I don't see a problem with
allowing it to be a function reference that returns a procedure pointer,
including the case of a defined operation -- there's no syntax ambiguity
here, as there was in the case of the READ statement.

Type-bound procedures (not references to them) would be useful, but
perhaps a different and more difficult story.  I think we already had a
protracted discussion about them being actual arguments.