client_handler.cpp exposes all the things I've been hinting at.  Pay special attention to the decision made in open() as well as the bit of cleverness in svc().

// $Id: client_handler.cpp,v 1.1 1998/08/30 23:47:13 schmidt Exp $

/*
   In client_handler.h I alluded to the fact that we'll mess around with a
   Client_Acceptor pointer.  To do so, we need the Client_Acceptor object
   declaration.

   We know that including client_handler.h is redundant because
   client_acceptor.h includes it.  Still, the sentry prevents double-inclusion
   from causing problems and it's sometimes good to be explicit about what
   we're using.

   On the other hand, we don't directly include any ACE header files here.
 */
#include "client_acceptor.h"
#include "client_handler.h"

/*
   Our constructor doesn't do anything.  That's generally a good idea.  Unless
   you want to start throwing exceptions, there isn't a really good way to
   indicate that a constructor has failed.  If I had my way, I'd have a boolean
   return code from it that would cause new to return 0 if I failed.  Oh
   well...
 */
Client_Handler::Client_Handler (void)
{
}

/*
   Our destructor doesn't do anything either.  That is also by design.
   Remember, we really want folks to use destroy() to get rid of us.  If that's
   so, then there's nothing left to do when the destructor gets invoked.
 */
Client_Handler::~Client_Handler (void)
{
}

/*
   The much talked about destroy() method!  The reason I keep going on about
   this is because it's just a Bad Idea (TM) to do real work inside of a
   destructor.   Although this method is void, it really should return
   int so that it can tell the caller there was a problem.  Even as
   void you could at least throw an exception which you would never want
   to do in a destructor.
 */
void Client_Handler::destroy (void)
{
  /*
     We probably got here because of an error on the stream, so the peer
     connection is probably already closed.  Still... there are other ways to
     get here and closing a closed peer doesn't hurt.
   */
  this->peer ().close ();

  /*
     Tell the reactor to forget all about us.  Notice that we use the same args
     here that we use in the open() method to register ourselves.  In addition,
     we use the DONT_CALL flag to prevent handle_close() being called.  Since we
     likely got here due to handle_close(), that could cause a bit of nasty
     recursion!
   */
  this->reactor ()->remove_handler (this,
        ACE_Event_Handler:: READ_MASK | ACE_Event_Handler::DONT_CALL);

  /*
     This is how we're able to tell folks not to use delete.  By
     deleting our own instance, we take care of memory leaks after ensuring
     that the object is shut down correctly.
   */
  delete this;
}

/*
   As mentioned before, the open() method is called by the Client_Acceptor when
   a new client connection has been accepted.  The Client_Acceptor instance
   pointer is cast to a void* and given to us here.  We'll use that to avoid
   some global data...
 */
int Client_Handler::open (void *_acceptor)
{
  /*
     We need this to store the address of the client that we are now connected
     to.  We'll use it later to display a debug message.
   */
  ACE_INET_Addr addr;

  /*
     Our ACE_Svc_Handler baseclass gives us the peer() method as a way to
     access our underlying ACE_SOCK_Stream.  On that object, we can invoke the
     get_remote_addr() method to get get an ACE_INET_Addr having our client's
     address information. As with most ACE methods, we'll get back (and return)
     a -1 if there was any kind of error.  Once we have the ACE_INET_Addr, we
     can query it to find out the clien's host name, TCP/IP address, TCP/IP
     port value and so forth.  One word of warning:   the get_host_name()
     method of ACE_INET_Addr may return you an empty string if your name server
     can't resolve it.  On the other hand, get_host_addr() will always give you
     the dotted-decimal string representing the TCP/IP address.
   */
  if (this->peer ().get_remote_addr (addr) == -1)
    {
      return -1;
    }

  /*
     Convert the void* to a Client_Acceptor*.  You should probably use those
     fancy new C++ cast operators but I can never remember how/when to do so.
     Since you can cast just about anything around a void* without compiler
     warnings be very sure of what you're doing when you do this kind of thing.
     That's where the new-style cast operators can save you.
   */
  Client_Acceptor *acceptor = (Client_Acceptor *) _acceptor;

  /*
     Our Client_Acceptor is constructed with a concurrency strategy.  Here, we
     go back to it to find out what that strategy was.  If thread-per-connection
     was selected then we simply activate a thread for ourselves and exit.  Our
     svc() method will then begin executing in that thread.

     If we are told to use the single-threaded strategy, there is no difference
     between this and the Tutorial 5 implementation.
   */
  if( acceptor->thread_per_connection() )
  {
    return this->activate();
  }

  /*
     Our reactor reference will be set when we register ourselves but I decided
     to go ahead and set it here.  No good reason really...
   */
  this->reactor (acceptor->reactor ());

  /*
     If we managed to get the client's address then we're connected to a real
     and valid client.  I suppose that in some cases, the client may connect
     and disconnect so quickly that it is invalid by the time we get here. In
     any case, the test above should always be done to ensure that the
     connection is worth keeping.

     Now, regiser ourselves with a reactor and tell that reactor that we want
     to be notified when there is something to read.  Remember, we took our
     reactor value from the acceptor which created us in the first place.
     Since we're exploring a single-threaded implementation, this is the
     correct thing to do.
   */
  if (this->reactor ()->register_handler (this, ACE_Event_Handler::READ_MASK) == -1)
    {
      ACE_ERROR_RETURN ((LM_ERROR, "(%P|%t) can't register with reactor\n"), -1);
    }

  /*
     Here, we use the ACE_INET_Addr object to print a message with the name of
     the client we're connected to.  Again, it is possible that you'll get an
     empty string for the host name if your DNS isn't configured correctly or
     if there is some other reason that a TCP/IP addreess cannot be converted
     into a host name.
   */
  ACE_DEBUG ((LM_DEBUG, "(%P|%t) connected with %s\n", addr.get_host_name ()));

  /*
     Always return zero on success.
   */
  return 0;
}

/*
   As mentioned in the header, the typical way to close an object in a threaded
   context is to invoke it's close() method.  Since we already have a handle_close()
   method built to cleanup after us, we'll just forward the request on to that
   object.
 */
int Client_Handler::close(u_long flags)
{
        this->handle_close(ACE_INVALID_HANDLE,0);

        /*
           After we've taken care of ourselves, call the baseclass method
           to do any other necessary cleanup.
         */
        return inherited::close();
}

/*
   In the open() method, we registered with the reactor and requested to be
   notified when there is data to be read.  When the reactor sees that activity
   it will invoke this handle_input() method on us.  As I mentioned, the _handle
   parameter isn't useful to us but it narrows the list of methods the reactor
   has to worry about and the list of possible virtual functions we would have
   to override.
 */
int Client_Handler::handle_input (ACE_HANDLE _handle)
{
  /*
     Some compilers don't like it when you fail to use a parameter.  This macro
     will keep 'em quiet for you.
   */
  ACE_UNUSED_ARG (_handle);

  /*
     Now, we create and initialize a buffer for receiving the data.  Since this
     is just a simple test app, we'll use a small buffer size.
   */
  char buf[128];
  memset (buf, 0, sizeof (buf));

  /*
     Invoke the process() method with a pointer to our data area.  We'll let
     that method worry about interfacing with the data.  You might choose to go
     ahead and read the data and then pass the result to process().  However,
     application logic may require that you read a few bytes to determine what
     else to read...  It's best if we push that all into the application-logic
     level.
   */
  return this->process (buf, sizeof (buf));
}

/*
   If we return -1 out of handle_input() or if the reactor sees other problems
   with us then handle_close() will be called.  It uses our destroy() method to
   shut us down cleanly and get rid of our instance.
 */
int Client_Handler::handle_close (ACE_HANDLE _handle, ACE_Reactor_Mask _mask)
{
  ACE_UNUSED_ARG (_handle);
  ACE_UNUSED_ARG (_mask);

  this->destroy ();
  return 0;
}

/*
   The ACE_Svc_Handler<> is ultimately derived from ACE_Task<>.  If you want to
   create a multi-threaded application, these are your tools!  Simply override
   the svc() method in your derivative and arrange for your activate() method
   to be called.  The svc() method then executes in the new thread.
 */
int Client_Handler::svc(void)
{
  /*
     Like handle_input(), we create a buffer for loading the data.  Doing so
     in handle_input() doesn't help any but there is a small performance increase
     by doing this here:  the buffer is created once when the thread is created
     instead of for each invocation of process().
   */
  char buf[128];

  // Forever...
  while( 1 )
  {
     // Clean the buffer...
     memset (buf, 0, sizeof (buf));

     /*
        Invoke the proces() method to read and process the data.  This is
        exactly the way it is used by handle_input().  That's the reason I
        created process() in the first place:  so that it can be used in either
        concurrency strategy.  Since process() has all of our application-level
        logic, it's nice that it doesn't have to change when we decide to go
        multi-threaded.

        Notice that since the recv() method call in process() blocks until
        there is data ready, this thread doesn't consume any CPU time until
        there is actually data sent from the client.
     */
     if( this->process(buf,sizeof(buf)) == -1 )
     {
       return(-1);
     }
  }

  return(0);
}

/*
   And, at last, we get to the application-logic level.  Out of everything
   we've done so far, this is the only thing that really has anything to do
   with what your application will do.  In this method we will read and process
   the client's data.  In a real appliation, you will probably have a bit more
   in main() to deal with command line options but after that point, all of the
   action takes place here.
 */
int Client_Handler::process (char *_rdbuf, int _rdbuf_len)
{
  /*
     Using the buffer provided for us, we read the data from the client. If
     there is a read error (eg -- recv() returns -1) then it's a pretty good
     bet that the connection is gone.  Likewise, if we read zero bytes then
     something wrong has happened.  The reactor wouldn't have called us if
     there wasn't some kind of read activity but there wouldn't be activity if
     there were no bytes to read...

     On the other hand, if we got some data then we can display it in a  debug
     message for everyone to see.
   */
  switch (this->peer ().recv (_rdbuf, _rdbuf_len))
    {
    case -1:
      ACE_ERROR_RETURN ((LM_ERROR, "(%P|%t) %p bad read\n", "client"), -1);
    case 0:
      ACE_ERROR_RETURN ((LM_ERROR, "(%P|%t) closing daemon (fd = %d)\n", this->get_handle ()), -1);
    default:
      ACE_DEBUG ((LM_DEBUG, "(%P|%t) from client: %s", _rdbuf));
    }

  return 0;
}

Well, that's it!  After all the talk & the hype, you would have expected it to be more difficult to create a multi-threaded server.  Surprise!  It really is that easy.  You still have to handle contention issues which we haven't addressed here and that is a rather nasty topic.  Still, for the simple case, this is all you have to do.

The next page is the last for this tutorial.  Head on over there & we'll round up the file list one last time.