Something new this time is client_acceptor.cpp.  I finally had enough code to move it out of the header.

// page04.html,v 1.10 2000/03/19 20:09:23 jcej Exp

#include "client_acceptor.h"

/* Construct ourselves with the chosen concurrency strategy.  Notice
   that we also set our Thread_Pool reference to our private instance.  */
Client_Acceptor::Client_Acceptor (int concurrency)
  : concurrency_ (concurrency),
    the_thread_pool_ (private_thread_pool_)
{
}

/* Construct ourselves with a reference to somebody else' Thread_Pool.
   Obvioulsy our concurrency strategy is "thread_pool_" at this point.  */
Client_Acceptor::Client_Acceptor (Thread_Pool &thread_pool)
  : concurrency_ (thread_pool_),
    the_thread_pool_ (thread_pool)
{
}

/* When we're destructed, we may need to cleanup after ourselves.  If
   we're running with a thread pool that we own, it is up to us to
   close it down.  */
Client_Acceptor::~Client_Acceptor (void)
{
  if (this->concurrency() == thread_pool_ && thread_pool_is_private ())
    thread_pool ()->close ();
}

/* Similar to the destructor (and close() below) it is necessary for
   us to open the thread pool in some circumstances.

   Notice how we delegate most of the open() work to the open() method
   of our baseclass.  */
int
Client_Acceptor::open (const ACE_INET_Addr &addr,
                       ACE_Reactor *reactor,
                       int pool_size)
{
  if (this->concurrency() == thread_pool_ && thread_pool_is_private ())
    thread_pool ()->open (pool_size);

  return inherited::open (addr, reactor);
}

/* Here again we find that we have to manage the thread pool.  Like
   open() we also delegate the other work to our baseclass.  */
int
Client_Acceptor::close (void)
{
  if (this->concurrency() == thread_pool_ && thread_pool_is_private ())
    thread_pool ()->close ();

  return inherited::close ();
}

Nothing really surprising here.  Most of it just manages the Thread_Pool.