| C & C++ Programming | MFC Programming | Download | Site Index |



Supplementary Note:

MFC Automation and Remote Automation




What do we have in this Module?

  1. Intro

  2. Passing Parameters in Automation

  3. Remote Automation

  4. History of DCOM

  5. Where Does Remote Automation Fit In?

  6. What Does Remote Automation Provide?

  7. Security in Remote Automation

  8. Remote Automation Threading Models

  9. Remote Automation Installation

  10. The Remote Automation Connection Manager

  11. Remote Automation User Components

  12. Creating Programs That Use Remote Automation

  13. To execute the Remote Automation object





Automation (formerly known as OLE Automation) makes it possible for one application to manipulate objects implemented in another application, or to expose objects so they can be manipulated.

An Automation server is an application (a type of COM server) that exposes its functionality through COM interfaces to other applications, called Automation clients. The exposure enables Automation clients to automate certain functions by directly accessing objects and using the services they provide.

Automation servers and clients use COM interfaces that are always derived from IDispatch and take and return a specific set of data types called Automation types. You can automate any object that exposes an Automation interface, providing methods and properties that you can access from other applications. Automation is available for both OLE and COM objects. The automated object might be local or remote (on another machine accessible across a network); therefore there are two categories of automation:

  • Automation (local).

  • Remote Automation (over a network, using Distributed COM, or DCOM).

Exposing objects is beneficial when applications provide functionality useful to other applications. For example, an ActiveX control is a type of Automation server; the application hosting the ActiveX control is the automation client of that control.

As another example, a word processor might expose its spell-checking functionality to other programs. Exposure of objects enables vendors to improve their applications by using the ready-made functionality of other applications. In this way, Automation applies some of the principles of object-oriented programming, such as reusability and encapsulation, at the level of applications themselves.

More important is the support Automation provides to users and solution providers. By exposing application functionality through a common, well-defined interface, Automation makes it possible to build comprehensive solutions in a single general programming language, such as Microsoft Visual Basic, instead of in diverse application-specific macro languages.

Many commercial applications, such as Microsoft Excel and Microsoft Visual C++, allow you to automate much of their functionality. For example, in Visual C++, you can write VBScript macros to automate builds, aspects of code editing, or debugging tasks.


Passing Parameters in Automation


One difficulty in creating Automation methods is providing a uniform "safe" mechanism to pass data between automation servers and clients. Automation uses the VARIANT type to pass data. The VARIANT type is a tagged union. It has a data member for the value (this is an anonymous C++ union) and a data member indicating the type of information stored in the union. The VARIANT type supports a number of standard data types: 2- and 4-byte integers, 4- and 8-byte floating-point numbers, strings, and Boolean values. In addition, it supports the HRESULT (OLE error codes), CURRENCY (a fixed-point numeric type), and DATE (absolute date and time) types, as well as pointers to IUnknown and IDispatch interfaces.

The VARIANT type is encapsulated in the COleVariant class. The supporting CURRENCY and DATE classes are encapsulated in the COleCurrency and COleDateTime classes.


Remote Automation


It is recommended that Visual C++ .NET developers use DCOM rather than Remote Automation for new applications. Visual C++ .NET does not support Windows 95. Remote Automation is a type of Automation that allows an interface consumer to execute an interface provider that resides on another machine, for example, on a network. The following sections explain how to create Automation objects that can be invoked and executed remotely, and how to create Automation controllers that can use these Remote Automation objects. It also examines configuration options and points out the major differences between Remote Automation and DCOM (the distributed version of COM and OLE that allows interfaces other than those related to automation to be invoked and executed remotely).


History of DCOM


When Automation was first introduced in early 1993, it was capable of being used only between applications running on the same machine. However, because it shared the same underpinnings as the rest of OLE, that is, COM (or Component Object Model), it was always intended that it would become "remotable" when COM itself was updated to include remoting capabilities. It was also planned that the transition from purely local operation to distributed operation would require little or no change to existing code.

So what does "remoting" mean? Local COM dictated that the consumer of an interface reside and execute on the same machine as the provider of that interface. For example, Microsoft Visual Basic could control a copy of Microsoft Excel on your desktop machine, but it was not capable of directing the execution of Excel on another machine. With the development of distributed COM, the consumer of an interface no longer needs to reside on the same machine as that on which the interface provider executes.

Once COM was adapted to work across a network, then any interface that was not tied to a local execution model (some interfaces have inherent reliance on local machine facilities, such as those drawing interfaces whose methods have handles to device contexts as parameters) would have the capability of being distributed. An interface consumer would make a request for a given interface; that interface may be provided by an instance of an object running (or to be run) on a different machine. The distribution mechanism inside COM would connect the consumer to the provider in such a way that method calls made by the consumer would appear at the provider end, where they would be executed. Any return values would then be sent back to the consumer. To all intents and purposes, the act of distribution is transparent to both the consumer and the provider.

Such a variety of COM does now exist. DCOM (for "distributed COM") has shipped with versions of Windows NT beginning with version 4.0 and including Windows 2000. Since late 1996, it has also been available for Windows 9x. In both cases, DCOM comprises a set of replacement and additional DLLs, with some utilities, which provide both local and remote COM capabilities. It is therefore now an inherent part of Win32-based platforms, and will be made available on other platforms by other organizations over time.





Where Does Remote Automation Fit In?


DCOM was released 1996 and is available with 32-bit platforms only. The Visual Basic team at Microsoft has always seen Visual Basic as using Automation to allow its components to communicate. The lack of a distributed version severely limited the use of these capabilities in enterprise environments, so the team developing Visual Basic 4.0 Enterprise Edition decided to investigate the creation of its own set of remoting components for the Automation parts of OLE and COM. Clearly, a major goal was to ensure that the result would be compatible with and could be replaced by DCOM when it became available. They then proceeded to implement Remote Automation (RA) for both 16-bit and 32-bit Windows platforms.

Remote Automation is not tied to any specific language, but until the release of Visual C++ 4.2 Enterprise Edition, it was shipped only with Visual Basic 4.0. Be aware that Remote Automation is wholly subsumed by DCOM. If you have the opportunity to use DCOM instead of Remote Automation in your applications, you should do so. Nevertheless, there are scenarios where Remote Automation is more appropriate:

There need be no difference between programs created to use Remote Automation and those created to use Automation over DCOM, and the configuration utilities make it very simple to switch operation between Remote Automation and DCOM. Consequently, it is not difficult to upgrade an application from Remote Automation to DCOM once the infrastructure is in place.


What Does Remote Automation Provide?


Remote Automation allows programs to invoke IDispatch implementations on one machine from another. It also supports other interfaces required by Automation, specifically IEnumVARIANT for collection support. It does not provide the ability to distribute any other COM interface (except IUnknown, of course) and, like regular Automation, it contains marshaling support only for those data types supported by Automation.

This set of facilities allows a program to access the methods and properties, including those that return collections or further automation objects, of an object running on an accessible network node. If the client machine is also running the appropriate software, it is possible for the server to call back to the client, again using Automation facilities (this works for 32-bit clients only, and is conceptually similar to events, although it does not use the same mechanism). For an application to be operable as a Remote Automation server, it must be implemented as an executable (that is, as a "local server" rather than as an "inproc server").


Security in Remote Automation


Remote Automation supports a basic level of security to allow a server application writer (or, rather, its administrator) to specify how a specific object may be activated remotely. All automation objects on a given system may be globally set to "disallow remote activation" or to "allow remote activation". Additionally, and more often, individual objects may be given such capabilities. Remote Automation uses a key in each object's registry settings, AllowRemoteActivation, to determine whether a given server may be activated remotely. If the systemwide settings use this mode, then each object in the registry may be assigned this key, and the individual status of each one may be set to "yes" or "no" as appropriate.

If the server system is running Windows NT or Windows 2000, then an alternative form of security is allowed. In this case, Remote Automation uses the NT access control list (ACL) to specify which users or group or groups of users may remotely activate a given server. Note that the security options apply to the whole object; it is not possible to set attributes of a specific interface or of individual properties or methods on that object. All security options may be set through the Remote Automation Connection (RAC) Manager.


Remote Automation Threading Models


Remote Automation supports both the single-threaded and the apartment-threaded models of COM. On 32-bit platforms, which are all that Visual C++ supports, the threading model should be set to "apartment".


Remote Automation Installation


Remote Automation has relatively few components:

Of these, RAC Manager is written in Visual Basic and therefore needs the Visual Basic run-time support. These and the other Remote Automation files are installed by Setup when you install Visual C++ Enterprise Edition. If you copy the Remote Automation components to a computer on which Visual C++ version Enterprise Edition is not installed, ensure that REGSRV32.EXE is on the computer's path, and register RACREG32.DLL using the following command line:




Note:   Versions of RAC Manager before Visual C++ 5.0 required GUAGE32.OCX and TABCTL32.OCX. Neither of these is required for the version of RAC Manager that ships with Visual C++ Enterprise Edition, version 5.0 or later.


The Remote Automation Connection Manager


To configure both client and server machines, you need to make registry changes. Rather than doing this by hand, it is far easier to use the Remote Automation Connection (RAC) Manager tool. This tool, RACMGR32.EXE, along with RACREG32.DLL, needs to be copied to any directory you choose. By putting it in the PATH, it can be executed from the taskbar using Run. Alternatively, you can create a shortcut to it or place a reference to it on the Start menu. RACMGR32 is written in Visual Basic and therefore needs some Visual Basic support DLLs. These files are placed in the same directory as RACMGR32.EXE on the CD-ROM. The versions of these files that are installed by the Setup for Visual C++ Enterprise Edition are equivalent or more recent than those that shipped with Visual Basic Enterprise Edition 5.0. The Visual C++ Setup copies the new versions of the Visual Basic files to your system directory. For Windows 9x, this directory is typically C:\Windows\System. For Windows NT and Windows 2000, it is typically C:\WINNT\system32. Setup also registers these files with the operating system. You may remove them from your Visual Basic installation.


Remote Automation User Components


You will need to ensure that each client machine contains your client program and any support DLLs it requires. You will also need to ensure that the server application and any support DLLs it requires are present on the server machine. Finally, you will need to ensure that your server program is registered on each client machine before RAC Manager can be run to configure the connection. If the program is self-registering (as most will be), you need only execute the server program on the client machine to register it. Failing that, you may have to execute a registration file that you provide, or manually edit the registry.


Creating Programs That Use Remote Automation


Any automation object, and any automation controller, is able to use Remote Automation without any change to the source code, without the need for recompilation, and without the need for relinking. Once you have a setup that works locally (that is, on the same machine), you need go through only a few steps to execute it remotely.


To execute the Remote Automation object...


  1. Register the application on the client machine or machines.

  2. Configure the client access to use remote server.

  3. Install and register the application on the server machine or machines.

  4. Configure the server to allow remote activation.

  5. Install the Automation Manager on the server machine(s).

  6. Run Automation Manager on the server(s).

  7. Run the application on the client(s).


Step 1 is most easily accomplished by loading and executing the server application on the client machine, as most servers are self registering. If you tested the setup locally beforehand, this stage is already complete. If the server application is not self registering, you may want to make it so. Otherwise, you will need to provide a registration file that the local user can run to perform this step. If you do neither of these things, you or an administrator will need to edit the registry manually, a procedure which is not recommended for all but the most advanced users.

Step 2 involves the use of the Remote Automation Connection (RAC) Manager. Run RAC Manager and ensure that the server connection tab is uppermost. Next, find the entry for the server object in the OLE Classes pane and click on it. Then move to the Network Address combo box and enter the name of the server machine on which the remote executable file will be run. For example, you may enter \\MyServer here. Then choose the appropriate network protocol from the list provided. You may need to check with your network administrator to determine which protocol is recommended. In many cases, this will be TCP/IP. Finally, you may want to choose an authentication level. In most cases, this will be left as (None) or Default. Now right-click the server in the OLE Classes pane. This will produce a shortcut menu from which you can select local or remote operation. Select remote.

Step 3 involves properly installing and registering the server application on the selected server machine or machines. Again, if the application is self registering, executing it once will also register it.

Step 4 involves configuring the server to allow remote execution. Run RAC Manager on the server machine, and ensure that the Client Access tab has the focus. Choose the activation model that you want (typically Allow Remote Creates by Key. If you do choose this option, you also need to click the Allow Remote Activation check box to set the value of the registry entry to 'Y'). If you are running Windows NT or Windows 2000 and you choose the Allow Remote Creates (ACL) option, you also have the option to edit the ACL by pushing the Edit ACL button.

To allow Remote Automation to work, you then need to ensure that the Automation Manager is installed and running on the server machine or machines. If it is not installed, copy AUTMGR32.EXE to the Windows system directory. To start Remote Automation, execute the Automation Manager. It will display a small status window in which a number of messages will be shown. Once it has started, it will minimize itself. If you want to continue to see status information, you can click the Automation Manager tab in the task bar to restore the window.

The final step is to execute the client application on the client machine. If you have followed the steps above, it should connect to the server object and execute precisely as it did locally, albeit a little slower. Notice that the RAC Manager also allows you to switch between Remote Automation and distributed COM (DCOM, on those platforms that support DCOM) with a single click of a radio button. If you choose DCOM, you can set various other configuration options.



| C & C++ Programming | MFC Programming | Download | Site Index |