Did anyone see this message or was I wasting my breath?

In article <NWILSON.94Aug19143245@lnis116.is.morgan.com> nwilson@lnis116.is.morgan.com (Ab Wilson) writes:

   Newsgroups: comp.std.c++
   Path: is1.is.morgan.com!is0.is.morgan.com!nwilson
   From: nwilson@lnis116.is.morgan.com (Ab Wilson)
   Sender: news@is.morgan.com
   Nntp-Posting-Host: lnis116
   Organization: /u/nwilson/.organization
   Distribution: World
   Date: Fri, 19 Aug 1994 13:32:45 GMT
   Lines: 52


   Are the following issues addressed in ANSI?

   At present this code isn't legal (well not with any compiler I've
   used).

   class Foo;

   void f(Foo*);

   typedef SomeTemplate<SomeType> Foo;

   or

   enum Foo { foo, bar };

   Why should I have to know what kind of type  I'm predeclaring? What if
   I change the implementation at some later date? I think the above code
   should be legel since it would reduce the ammount of coupling required
   between header files.

   Another proposal...

   If I have

   class Wibble
   {
   public:
     virtual void wobble();
   };

   and wobble is  never defined, I can't link  an executable that uses  a
   Wibble  even if wobble is  never used. (This  is because Wibble's vtbl
   has to be filled in with the address of wobble.)

   For a  real  instance of where this   becomes a pain imagine  a system
   developed from two libraries.   One library deals with the application
   types (models), the other deals with GUI  code (views). I'd like to be
   able to  keep the  two libraries separate  so  that  I can  build test
   programs  that don't have  to link with  the GUI code.  However I also
   want my models to  have a virtual   makeView method. I would  like the
   standard to specify  that the linker should  not complain about wobble
   being undefined.

   Both these issues affect real developers. Neither would break existing
   code or  add new constructs to  the language. If I've missed something
   and these issues have already been addressed then please forgive me.

   Could any  followups be  directed to me  personally  since I am not  a
   regular reader of this group.

    Ab.

Are the following issues addressed in ANSI?

At present this code isn't legal (well not with any compiler I've
used).

class Foo;

void f(Foo*);

typedef SomeTemplate<SomeType> Foo;

or

enum Foo { foo, bar };

Why should I have to know what kind of type  I'm predeclaring? What if
I change the implementation at some later date? I think the above code
should be legel since it would reduce the ammount of coupling required
between header files.

Another proposal...

If I have

class Wibble
{
public:
  virtual void wobble();
};

and wobble is  never defined, I can't link  an executable that uses  a
Wibble  even if wobble is  never used. (This  is because Wibble's vtbl
has to be filled in with the address of wobble.)

For a  real  instance of where this   becomes a pain imagine  a system
developed from two libraries.   One library deals with the application
types (models), the other deals with GUI  code (views). I'd like to be
able to  keep the  two libraries separate  so  that  I can  build test
programs  that don't have  to link with  the GUI code.  However I also
want my models to  have a virtual   makeView method. I would  like the
standard to specify  that the linker should  not complain about wobble
being undefined.

Both these issues affect real developers. Neither would break existing
code or  add new constructs to  the language. If I've missed something
and these issues have already been addressed then please forgive me.

Could any  followups be  directed to me  personally  since I am not  a
regular reader of this group.

 Ab.