What's required to create good software? While it's possible to write first-rate code in almost any environment, creating good software is much easier when the right platform and tools are available. For most Windows developers today, that platform is defined by .NET. While defining .NET clearly was once a challenge, it's now clear that the .NET label refers primarily to two things. They are:

  • The .NET Framework, which consists of the Common Language Runtime (CLR) and the .NET Framework class library. The CLR provides a standard foundation for building applications, while the .NET Framework class library offers a large set of standard classes and other types that can be used by any .NET Framework application written in any language.
  • Visual Studio, an integrated development environment (IDE) for creating Windows applications. While this tool can be used to build software that runs directly on Windows, its main focus is helping developers create .NET Framework applications. Visual Studio supports several programming languages for creating these applications, including C#1, Visual Basic (VB), and C++.

Note - The .NET Framework and Visual Studio are the main components of .NET

Perspective: .NET's Naming Journey - It makes sense today to think of the name ".NET" as primarily referring to the .NET Framework and Visual Studio. Things weren't always so simple, however. When .NET was first announced in the summer of 2000, Microsoft applied the term to a broad range of things. Today's .NET technologies were included, of course, but so were several other things. Many of Microsoft's server products, including SQL Server and BizTalk Server, were grouped together as the .NET Enterprise Servers, for example, and a wholly separate effort eventually known as .NET My Services was launched. There was even talk about a possible Windows .NET and Office .NET sometime in the future.

But was there a common technical underpinning for all of these things? Sadly, the answer was no. When Microsoft first sprang .NET on the world, it treated the term as a broad brand, one that could be applied to pretty much anything the company was doing. The result was a good deal of confusion among Microsoft's customers.

Thankfully, the story has gotten much simpler. The .NET Enterprise Servers are now considered part of the Windows Server System, and so they've lost the .NET tag. .NET My Services faded from the scene, while the branding boffins in Redmond decided against tacking the .NET brand onto either Windows or Office. Today, when somebody says ".NET," they're referring to the .NET Framework and Visual Studio.

Various versions exist for both of these technologies. The versions described in this book are those released by Microsoft in late 2005: version 2.0 of the .NET Framework and Visual Studio 2005.

The .NET Framework

The heart of .NET is the .NET Framework. First released in 2002, it brought enormous change to the lives of those who write Windows software and the people who manage them. Figure 1-1 shows the Framework's two main parts: the CLR and the .NET Framework class library. A .NET application always uses the CLR, and it can also use whatever parts of the class library it requires.

Figure 1-1 - The .NET Framework consists of the Common Language Runtime (CLR) and the .NET Framework class library.

Every application written using the Framework depends on the CLR. Among other things, the CLR provides a common set of data types, acting as a foundation for C#, VB, and all other languages that target the .NET Framework. Because this foundation is the same no matter which language they choose, developers see a more consistent environment.

Note - The CLR provides a common basis for all .NET languages

Perspective: The .NET Framework vs. the Java Environment

Mainstream software development today has split largely into two camps. Microsoft, promoting the .NET Framework, is in one, while most other vendors backing the Java environment are in the other. Each technology has its fans and detractors, and each has a substantial installed base today.

These competing worlds are strikingly similar. To see how similar, compare the figure above with Figure 1-1. Both environments support the same kinds of applications, and both provide a large standard library to help build those applications. The Java library, mostly known today as Java 2 Enterprise Edition (J2EE or just JEE) includes Java Server Pages (JSP) for Web scripting, Swing for building GUIs, JAX-WS (formerly JAX-RPC) for Web services-based communication, Enterprise JavaBeans (EJB) for building scalable server applications, JDBC for database access, and other classes. These technologies are analogous to the .NET Framework's ASP.NET, Windows Forms, ASP.NET Web Services, Enterprise Services, and ADO.NET, respectively. The Java virtual machine is also much like the .NET Framework's CLR, and even the semantics of the dominant languages-Microsoft's C# and VB vs. Java-are quite similar.

There are also differences, of course. One obvious distinction between the two is that the Java environment runs on diverse operating systems, while the .NET Framework focuses on Windows. The trade-off here is clear: Portability is good, but it prevents tight integration with any one system, and integration is also good. You can't have everything, at least not all at the same time. Also, Java-based products are available from multiple vendors, while only Microsoft provides the .NET Framework. Different Java vendors can provide different extensions to the core specifications, so developers can get somewhat locked into a single vendor. Still, portability across different Java platforms is possible, while the .NET Framework unambiguously ties your application to Microsoft.

This bifurcation and the competition it engenders are ultimately a good thing. Both camps have had good ideas, and each has borrowed from the other. Having one completely dominant technology, whether the .NET Framework or Java, would produce a stultifying monopoly, while having a dozen viable choices would lead to anarchy. Two strong competitors, each working to outdo the other, is just right.

Applications written in any .NET language can use the code in the .NET Framework class library. Among the most important technologies provided in this library are the following:

  • ASP.NET: Classes focused on building browser-accessible applications.
  • Windows Forms: Classes for building Windows graphical user interfaces (GUIs) in any CLR-based programming language.
  • ASP.NET Web Services (also called ASMX): Classes for creating applications that communicate with other applications using Web services.
  • Enterprise Services: Classes that provide distributed transactions, object instance control, and other services useful for building reliable, scalable applications.
  • ADO.NET: Classes focused on accessing data stored in relational database management systems (DBMS).

The .NET Framework class library contains much more than this short list indicates. Among the other services it provides are support for creating and working with XML documents, a variety of security services, and mechanisms for interoperating with applications built using older Windows technologies such as the Component Object Model (COM).

Note - The .NET Framework class library provides standard code for common functions

As this short description suggests, the .NET Framework class library can be used to create many different types of applications. And because all of the services in this library are built on the CLR, applications can combine them as needed. A browser application built using ASP.NET, for example, might use ADO.NET to access stored data and Enterprise Services to perform distributed transactions.

Note - The .NET Framework supports various kinds of applications

Software that uses the .NET Framework (and thus relies on the CLR) is referred to as managed code. As Figure 1-2 shows, an application can be built solely from managed code, relying entirely on the CLR and the relevant parts of the .NET Framework class library. An application can also be built from a combination of managed code and ordinary unmanaged code, with the two interacting as necessary. This second option, shown on the right side of the figure, is especially important for existing applications. Most new Windows applications created today are built wholly in managed code, but it can also be useful to extend pre-.NET applications with managed code. And although it's not shown in the figure, it's still possible to create new Windows applications entirely in unmanaged code-using the .NET Framework isn't obligatory.

Figure 1-2 - An application can be built entirely from managed code or from a combination of managed and unmanaged code.

Note - A .NET Framework application consists of managed code

Managed code is typically object-oriented, so the objects it creates and uses are known as managed objects. A managed object can use and inherit from another managed object even if the two are written in different languages. This fact is a key part of what makes the .NET Framework class library an effective foundation: Objects written in any CLR-based language can inherit and use the class library's code. Given the fundamental role played by the CLR, understanding the .NET Framework begins with understanding this runtime environment.

Note - Managed code is typically built using managed objects

The Common Language Runtime

Built from scratch to support modern applications, the CLR embodies a current view of what a programming environment should be. While it's hard to claim complete originality for any idea in computer science today, it is fair to say that this essential .NET technology takes an interesting new approach to programming languages.

What the CLR Defines

Think about how a programming language is typically defined. Each language commonly has its own unique syntax, its own set of control structures, a unique set of data types, its own notions of how classes inherit from one another, and much more. The choices a language designer makes are driven by the target applications for the language, who its users are meant to be, and the designer's own sensibilities.

Note - There's widespread agreement on the features a modern programming language should offer

Yet most people agree on much of what a modern general-purpose programming language should provide. While opinions on syntax differ-some developers love curly braces, others abhor them-there's widespread agreement on what semantics a language should offer. Given this, why not define a standard implementation of those semantics, then allow different syntaxes to be used to express those semantics?

Looking Backward: Windows DNA and COM - For most of the 1990s, application developers in the Microsoft environment relied on a set of technologies that became known as Windows DNA. Those technologies included the Component Object Model (COM) and Distributed COM (DCOM), a larger group of COM-based technologies known collectively as COM+, support for creating browser applications using Active Server Pages (ASP), data access support with ActiveX Data Objects (ADO), and others. The most commonly used languages for building Windows DNA applications were VB and C++, both supported by earlier versions of Visual Studio.

Tens of thousands of applications based on these technologies are in production today, providing solid evidence of Windows DNA's success. Yet the technologies Windows DNA includes were developed independently over a period of several years. Because of this, the integration among them wasn't as complete as it might have been. For example, while the Windows DNA environment let developers use various programming languages, each language has its own runtime libraries, its own data types, its own approach to building GUIs, and other differences. Applications written in different languages also accessed system services in different ways. C++ applications could make direct calls to the operating system through the Win32 interface, for instance, while VB applications typically accessed these services indirectly. These differences made life challenging for developers working in more than one language. COM, by defining common conventions for interfaces, data types, and other aspects of interaction among different software, was effectively the duct tape that held this complex environment together.

By providing a common foundation that can be used from all languages, the .NET Framework significantly simplified life for Windows developers. Applications built on the .NET Framework don't face many of the problems that COM addresses-.NET Framework applications all use the CLR, for example, which defines a common approach to interfaces and other data types-and so the glue between different languages that COM provides is no longer necessary. This is why COM technology isn't used in building pure .NET Framework applications. Instead, developers can build software that interacts in a more natural and substantially simpler way.

For the most part, the arrival of the .NET Framework was the death knell for Windows DNA and COM. It's taken quite a while for organizations to migrate to the new world of .NET, and many applications built with these older Windows technologies are still in production. Still, with few exceptions, serious new Windows development today uses .NET, not Windows DNA.

Note - The CLR defines a common set of semantics that is used by multiple languages

The CLR provides this standard implementation. By providing a common set of data types such as integers, strings, classes, and interfaces, specifications for how inheritance works, and much more, it defines a common set of semantics for languages built on it. The CLR says nothing about syntax, however. How a language looks, whether it contains curly braces or semicolons or anything else, is entirely up to the language designer. While it is possible to implement languages with varying behaviors on top of the CLR, the CLR itself provides a consistent, modern set of semantics for a language designer to build on.

Note - The CLR also provides other common services

Along with its standard types, the CLR provides other fundamental services. Those services include the following:

  • Garbage collection, which automatically frees managed objects that are no longer referenced.
  • A standard format for metadata, information about each type that's stored with the compiled code for that type.
  • A common format, called assemblies, for organizing compiled code. An assembly can consist of one or more Dynamic Link Libraries (DLLs) and/or executables (EXEs), and it includes the metadata for the classes it contains. A single application might use code from one or more assemblies, and so each assembly can specify other assemblies on which it depends.

Using the CLR

The CLR was not defined with any particular programming language in mind. Instead, its features are derived largely from popular existing languages, such as C++, the pre-.NET version of VB, and Java. Today, Microsoft provides several CLR-based languages, including C#, the .NET version of VB, an extended version of C++, and others. Third parties also provide languages built on the CLR.

Note - The CLR supports many different programming languages

Perspective: Standardizing C# and the CLR - C# and a subset of the CLR called the Common Language Infrastructure (CLI) are now official international standards. Microsoft originally submitted them to the ECMA standards organization, and they've also been approved by the International Organization for Standardization (ISO). Along with C#, the standardized technologies include the syntax and semantics for metadata, MSIL (rechristened the Common Intermediate Language, or CIL), and parts of the .NET Framework class library. For more details on exactly what has been submitted and its current status, see http://msdn.microsoft.com/net/ecma.

The biggest effect of establishing C# and the CLI as standards is to let others more easily implement these technologies. The most visible non-Microsoft implementation of .NET is surely the Mono project (http://www.mono-project.com). Mono's ambitious goal is to implement at least a large part of what Microsoft has given to ECMA, including a C# compiler and the CLI, along with other parts of the .NET Framework. Mono's creators say that they were attracted to the CLR for technical reasons, which must please Microsoft. (In fact, it's possible to view the CLI as the specification of a system, while .NET's CLR is just the Microsoft implementation of this specification.) Now led by Novell, the Mono project qualifies as a success in many ways, with implementations available for Linux and other operating systems. If nothing else, I admire the ambition and ability of the people who created it.

No matter what language it's written in, all managed code is compiled into Microsoft Intermediate Language (MSIL) rather than a machine-specific binary. MSIL (also referred to as just IL) is a set of CPU-independent instructions for performing typical operations such as loading and storing information and calling methods. Each DLL and EXE in an assembly contains MSIL rather than processor-specific code. Installing a .NET Framework application on your system really means copying to your disk files that contain MSIL rather than a machine-specific binary. When the application is executed, MSIL is transformed into native code before it's executed.

Note - Managed code is always compiled first into MSIL

Figure 1-3 illustrates the process of compiling and executing managed code. Source code written in VB, C#, or another language that targets the CLR is first transformed into MSIL by the appropriate language compiler. As the figure shows, the compiler also produces metadata that's stored in the same file as the MSIL. Before execution, this MSIL is compiled into native code for the processor on which the code will run. By default, each method in a running application is compiled the first time that method is called. Because the method is compiled just in time to execute it, this approach is called just in-time (JIT) compilation.

Figure 1-3 - All managed code is compiled first to MSIL, then translated into native code before execution.

Note - Each method is typically JIT compiled the first time it's invoked

One point worth noting is that any language built on the CLR should exhibit roughly the same performance as any other CLR-based language. Unlike the pre-.NET world, where the performance difference between VB and C++ was sometimes significant, a .NET Framework application written in C# isn't noticeably faster than the same application written in VB. While some compilers may produce better MSIL code than others, large variations in execution speed are unlikely.

Note - All .NET Framework-based languages have about the same level of performance

The CLR is the foundation of everything else in the .NET Framework. All code in the .NET Framework class library depends on it, as do all Framework-based applications. Chapter 2 provides a more detailed look at the technology of the CLR.

The .NET Framework Class Library

The .NET Framework class library is exactly what its name suggests: a library of classes and other types that developers can use to make their lives easier. While these classes are themselves written in C#, they can be used from any CLR-based language. Code written in C#, VB, C++, or any other language supported by the .NET Framework can create instances of these classes and call their methods. That code can also rely on the CLR's support for inheritance to inherit from the library's classes.

Note - The .NET Framework class library can be used from any CLR-based language

Surveying the Library

The contents of the .NET Framework class library are organized into a tree of namespaces. Each namespace can contain types, such as classes and interfaces, and other namespaces. Figure 1-4 shows a very small part of the .NET Framework class library's namespace tree. The namespaces shown include the following:

Figure 1-4 - The .NET Framework class library is structured as a hierarchy of namespaces, with the System namespace at the root.
  • System: The root of the tree, this namespace contains all of the other namespaces in the .NET Framework class library. System also contains the core data types used by the CLR (and thus by languages built on the CLR). These types include several varieties of integers, a string type, and many more.
  • System.Web: This namespace contains types useful for creating Web applications, and like many namespaces, it has subordinate namespaces. Developers can use the types in System.Web.UI to build ASP.NET browser applications, for example, while those in System. Web.Services are used to build ASP.NET Web Services applications.
  • System.Data: The types in this namespace comprise ADO.NET. For example, the Connection class is used to establish connections to a database management system (DBMS), while an instance of the DataSet class can be used to cache and examine the results of a query issued against that DBMS.
  • System.Windows.Forms: The types in this namespace make up Windows Forms, and they're used to build Windows GUIs. Rather than relying on language-specific mechanisms, such as the older Microsoft Foundation Classes (MFC) in C++, .NET Framework applications written in any programming language use this common set of types to build graphical interfaces for Windows.
  • System.EnterpriseServices: The types in this namespace provide services required for some kinds of enterprise applications. Implemented by COM+ in the pre-NET world, these services include distributed transactions, object instance lifetime management, and more. The most important type in this namespace, one from which classes must inherit to use Enterprise Services, is the ServicedComponent class.
  • System.XML: Types in this namespace provide support for creating and working with XML-defined data. The XmlDocument class, for instance, allows accessing an XML document using the Document Object Model (DOM). This namespace also includes support for technologies such as the XML Schema definition language (XSD) and XPath.

Note - The .NET Framework class library is organized as a tree

The .NET Compact Framework - While the .NET Framework is useful for writing applications on desktops and server machines, it can also be used with smaller devices, such as mobile phones, PDAs, and set-top boxes. Small devices are becoming more and more important, and they're an important piece of Microsoft's overall business strategy. These devices typically have less memory, however, so they're unable to run the complete .NET Framework. The .NET Compact Framework addresses this issue. By eliminating some parts of the .NET Framework class library, it allows use of the Framework in smaller devices.

The .NET Compact Framework targets the Windows CE operating system, but because it's built on the same foundation used in larger systems, developers can use Visual Studio as their development environment. Organizations that must create software for a range of devices can now use the same languages, the same tools, and much of the same development platform to target systems of all sizes.

Many more namespaces are defined, providing support for file access, serializing an object's state, remote access to objects, and much more. In fact, the biggest task facing developers who wish to build on the .NET Framework is learning to use the many services that the library provides. There's no requirement to learn everything, however, so a developer is free to focus on only those things relevant to his or her world. Still, some parts will be relevant to almost everybody, and so the next sections provide a quick overview of some of this large library's most important aspects.

Note - Learning the .NET Framework class library takes time

Building Web Applications: ASP.NET

Implemented in the System.Web namespace, ASP.NET is an important piece of the .NET Framework. The successor to the very popular Active Server Pages (ASP) technology, ASP.NET applications are built from one or more pages. Each page contains HTML and/or executable code, and typically has the extension .aspx. As Figure 1-5 shows, a request from a browser made via HTTP causes a page to be loaded and executed. Any output the page creates is then returned to the browser that made the request.

Figure 1-5 - ASP.NET allows developers to create browser-accessible applications.

Note - ASP.NET applications rely on .aspx pages

Building effective Web applications requires more than just the ability to combine code with HTML. Accordingly, ASP.NET provides a range of support, including the following:

  • Web controls, allowing a developer to create a browser GUI in a familiar way. By dragging and dropping standard ASP.NET controls for buttons and other interface elements onto a form, it's possible to build GUIs for Web applications in much the same way as for local Windows applications.
  • Mechanisms for managing an application's state information.
  • Built-in support for maintaining information about an application's users, sometimes called membership information.
  • Support for data binding, which allows easier access to information stored in a DBMS or some other data source.

Note - ASP.NET includes a number of things to help developers create Web applications

Given the popularity of Web applications, ASP.NET probably impacts more developers than any other part of the .NET Framework class library. Chapter 5 provides more detail on this key component of the .NET Framework.

Accessing Data: ADO.NET

ADO.NET lets applications work with stored data. As Figure 1-6 shows, access to a DBMS relies on a .NET Framework data provider, written as managed code. Providers that allow access to SQL Server, Oracle, and other DBMS are included with the .NET Framework. They allow a client application to issue commands against the DBMS and examine any results those commands return. The result of a Structured Query Language (SQL) query, for example, can be examined in two ways. Applications that need only read the result a row at a time can do this by using a DataReader object to march through the result one record at a time. Applications that need to do more complex things with a query result, such as send it to a browser, update information, or store that information on disk, can instead have the query's result packaged inside a DataSet object.

Figure 1-6 - ADO.NET allows .NET Framework applications to access data stored in DBMS and XML documents.

Note - ADO.NET lets applications access stored data

As Figure 1-6 illustrates, a DataSet can contain one or more tables. Each table can hold the result of a different query, so a single DataSet might potentially contain the results of two or more queries, perhaps from different DBMS. In effect, a DataSet acts as an in-memory cache for data. As the figure shows, however, DataSets can hold more than just the result of a SQL query. It's also possible to read an XML document directly into a table in a DataSet without relying on a .NET Framework data provider. Data defined using XML has also become much more important in the last few years, so ADO.NET allows accessing it directly. While not all .NET Framework applications will rely on ADO.NET for data access, a large percentage surely will. ADO.NET is described in more detail in Chapter 6.

Note - An ADO.NET DataSet acts as an in-memory cache for data

Running the .NET Framework - The .NET Framework is meant to be the foundation for most Windows applications going forward. To make this possible, the Framework runs on many versions of Windows, including Windows 2000, Windows XP, Windows Server 2003, and Windows Vista. It's also available for the 64-bit versions of Windows XP, Windows Server 2003, and Windows Vista. The Framework doesn't run on older systems, however, such as Windows 95 or Windows NT. Given that it was released many years after these versions of Windows, this shouldn't be surprising.

The .NET Framework also supports an option called side-by-side execution. This allows simultaneous execution of not just multiple versions of the same application, but also multiple versions of the .NET Framework itself. For example, a single machine might have both version 1.1 and version 2.0 of the Framework installed, with each used to run applications written specifically for it. This lets organizations move forward with new versions of the .NET Framework without touching existing applications that run on earlier releases.

Building Distributed Applications

Creating software that communicates with other software is a standard part of modern application development. Yet different applications have different communication requirements. To meet these diverse needs, the .NET Framework class library includes three distinct technologies for creating distributed applications. Figure 1-7 illustrates these choices.

Figure 1-7 - Distributed applications can use ASP.NET Web Services, .NET Remoting, or Enterprise Services.

ASP.NET Web Services, mostly defined in System.Web.Services, allows applications to communicate using Web services. Since it's part of ASP.NET, this technology lets developers use a similar model for creating distributed software. As Figure 1-7 shows, applications that expose methods as Web services can be built from files with the extension .asmx, each of which contains only code. Clients make requests using the standard Web services protocol SOAP2, and the correct page is loaded and executed. Because this technology is part of ASP.NET, requests and replies also go through Internet Information Services (IIS), the standard Web server for Windows.

Note - ASP.NET Web Services allow communication via SOAP

Communication via Web services is especially useful for interoperating with software built on platforms other than the .NET Framework, such as the Java environment. But it's not always the right solution. In some situations, the technology known as .NET Remoting, defined in the System.Runtime.Remoting namespace, is a better choice. Unlike ASP.NET Web Services, .NET Remoting focuses on direct communication between applications built on the .NET Framework. While it does support a version of SOAP, this technology also provides a binary protocol along with the ability to add extensions using any other protocol a developer needs. .NET Remoting isn't the most common choice for communication, but it can be important for some kinds of applications.

Note - .NET Remoting focuses on communication between .NET Framework-based applications

-Perspective: The .NET Framework on Non-Windows Systems - Applications written using the .NET Framework are compiled to a processor-independent form-MSIL-and shield themselves from the vagaries of a specific operating system by writing to the .NET Framework class library. This is much like the Java world, where applications are compiled to bytecode and can rely on standard Java libraries rather than making direct calls to a specific operating system. Java was expressly designed to work on multiple processors and operating systems. Is the same thing true for .NET?

To some extent, the answer is clearly yes. Microsoft itself has provided a port of the Framework's fundamentals, known as the Shared Source CLI, to the FreeBSD version of UNIX. Theoretically, the .NET Framework could become a cross-platform solution on a wide range of systems.

Yet some technical issues remain. While MSIL is clearly platform independent, some parts of the .NET Framework class library just as clearly are not. Enterprise Services, for example, providing support for scalable, transaction-oriented applications, is based on an earlier Windows technology called COM+. Accordingly, this part of the library runs only where COM+ is available. Other parts of the class library also betray their Windows origins in more or less obvious ways.

Just as important, Microsoft would face challenges in making customers believe that it's serious about long-term support of the .NET Framework on non-Windows systems. The company's laser-like focus on its own operating systems has been a hallmark of its business, as well as a primary factor in its success. If Microsoft wished to make the .NET Framework a true multiplatform rival for Java, the technical potential is there. But so far, at least, .NET is fundamentally a Windows technology.

In fact, today's most visible non-Windows implementation of .NET, the Mono project, hasn't attracted much commercial attention. Some enterprises looking to build .NET applications had hoped that Mono would provide them with a safe, non-Microsoft alternative should they need one. In practice, though, this option hasn't been appealing to many organizations. Like it or not, the great majority of people who build software on the .NET Framework should expect to run that code on some version of Windows.

The third option for creating distributed applications using the .NET Framework is Enterprise Services. Defined in the System.EnterpriseServices namespace, it provides applications with services such as distributed transactions and more. Figure 1-7 illustrates this, showing a server application accessing two databases. If a single transaction included updates to both of these databases, Enterprise Services might well be the right choice for the application that used this transaction. Remote clients can communicate directly with an Enterprise Services application using DCOM, and it's also possible for an ASP.NET application to use Enterprise Services when necessary.

Note - Enterprise Services provides distributed transactions and other services

All three options make sense in different situations, and having a basic understanding of all three is useful. For a more detailed look at each of them, see Chapter 7.

Visual Studio 2005

Visual Studio is by far the most popular tool for creating .NET applications today. The current version, Visual Studio 2005, is the successor to Visual Studio .NET, which was itself the first version of this tool to support creating managed code. Both tools provide a very large set of services for developers, including all of the bells and whistles of a modern IDE.

Note - Most .NET developers today use Visual Studio

Figure 1-8 shows how Visual Studio looks to the creator of a simple .NET Framework application. Different windows show different aspects of the project, including the code itself and, in the lower right corner, any properties that may have been set. If you've never used a modern software development tool, don't be fooled by this very basic picture: Visual Studio is actually a complex and powerful IDE, one that's used by millions of developers today.

Figure 1-8 - Visual Studio provides a range of support for creators of .NET applications.

Visual Studio 2005 is actually a family of products, each aimed at a particular kind of developer. This family includes Visual Studio 2005 Express Edition, aimed at beginning developers and hobbyists, Visual Studio Standard Edition, aimed at more serious developers, and Visual Studio 2005 Professional Edition, intended for hard-core software professionals building scalable distributed applications. The product family also includes Visual Studio Tools for Office, which provides the ability to create applications in VB or C# that build on Excel, Word, Outlook, and InfoPath. Another important member of this family, one that's described in a bit more detail later in this chapter, is Visual Studio Team System, which provides a set of interrelated tools specialized for various roles in a development group.

Note - Visual Studio is a family of products

What's New in Visual Studio 2005 - The introduction of Visual Studio Team System and support for DSLs are probably the two biggest innovations in Visual Studio 2005. This new version of Microsoft's flagship development tool also adds several other interesting new features, including the following:

  • Refactoring support: Refactoring allows making small modifications to software that together can lead to significant improvement. Some common refactoring changes, such as extracting code and wrapping it in its own method, or encapsulating a field inside get and set methods, can be done through a built-in Visual Studio menu.
  • Edit and continue: In pre-.NET versions of Visual Basic, a developer could change code in a running application while it was being debugged, then continue execution from the point of the change. Known as Edit and Continue, plenty of people found this useful. Omitted from Visual Studio .NET (to the great concern of some), this feature is now available with all of the programming languages supported by Visual Studio 2005.
  • Code snippets: Many applications include code that performs common tasks. To make implementing these easier, Visual Studio 2005 provides snippets of prewritten code for common situations. Snippets are provided for declaring standard types, such as classes, creating common constructions, such as an if statement, and other typical programming tasks. Developers can also create new snippets then store them for future use.

As already described, several different programming languages can be used to create .NET Framework applications. Visual Studio 2005 ships with support for C#, VB, C++, and other languages, giving developers a range of choices. The 2005 version of the tool also adds support for using domain-specific languages (DSLs), an idea that's described later in this chapter. The main focus of Visual Studio, however, is helping developers create applications using general purpose CLR-based programming languages, and so the next section takes a brief look at this area.

Note - Visual Studio 2005 provides a variety of different languages

Perspective: The Fate of Pre-.NET Applications - Applications built using the Windows DNA technologies that preceded .NET, such as COM and Visual Basic 6 (VB6), won't generally run on the .NET Framework without at least some modification. This is a problem for the many organizations that have invested in these applications. But just how big a problem is it?

The answer depends on what kind of application we're talking about and what kind of organization is responsible for it. For example, think about a VB6 application built by a typical enterprise that solves a specific business problem. Installing the .NET Framework on the same machine that runs this app won't break the application, and since these older apps will happily run on newer versions of Windows, there's no requirement to change. If the application still meets the business need effectively, why invest the time and resources to rewrite it using .NET? True, the rewritten app would likely be better in some technical ways, but the benefits to its users-the people who ultimately pay for it-aren't likely to be large enough to justify the cost.

But suppose this VB6 application needs to be modified in some way. It's always possible to just keep on using an older version of Visual Studio, one that (unlike Visual Studio 2005) supports VB6. But Microsoft's support for these older development environments is fading away, and many organizations feel uncomfortable relying on unsupported tools. In this case, the application may need to be rewritten solely to avoid this fate.

What about applications written by independent software vendors (ISV)? For example, many ISVs (including Microsoft) have a large investment in applications written using pre-.NET versions of C++. Visual Studio 2005 allows working directly with raw C++-you don't have to use the CLR-and so there's no immediate need to change these applications. New Windows features are typically exposed via managed code, however, and so taking advantage of this new functionality will require extending the application with some CLR-based C++.

Perhaps the hardest problem is that faced by ISVs or enterprises with a substantial investment in a VB6 application that still has many years of use ahead of it. An application like this is likely to need substantial change over time, yet creating new functionality in VB6 means freezing ever more work into a legacy technology. One option is to create new code in the .NET version of VB, then connect these additions to what already exists. To make this possible, the CLR has built-in support that allows managed code to call existing DLLs, access and be accessed by COM objects, and interoperate in other ways with the previous generation of Windows software. The big challenge here is the problem just mentioned: Versions of Visual Studio that support VB6 are falling out of support (although the VB runtime itself is still supported). In this case, it might be worth the investment to rebuild the entire application on .NET, especially if significant new functionality can be added at the same time.

Yet another challenge of transitioning to a wholly new development environment is the cost of retraining developers. In the long run, avoiding the .NET Framework isn't possible for Windows-oriented organizations, so ponying up the cash for developer education is unavoidable. It's not cheap, and time will be lost as developers come up to speed on this new technology. The .NET Framework really is substantially better than its predecessors, however, so most organizations are likely to see productivity gains once developers have internalized this new environment.

General Purpose Languages

Although Visual Studio supports many different programming languages, the CLR defines the core semantics for most of them. Yet while the way those languages behave often has a good deal in common, a developer can still choose the language that feels most natural to her. All of these languages use at least some of the services provided by the CLR, and some expose virtually everything the CLR has to offer. Also, with the exception of C++, it's not possible to create traditional standalone binary executables in these languages. Instead, every application is compiled into MSIL and so requires the .NET Framework to execute. This section provides a short introduction to each of the most commonly used general purpose languages built on the CLR.

Note - Visual Studio 2005 supports CLR-based languages

C#

The two dominant languages for Windows development in the pre-.NET world were C++ and Visual Basic 6 (VB6). Both had sizable user populations (although VB6's user base was much larger), and so Microsoft needed to find a way to make both groups as happy as possible with their new environment. How, for example, could the large number of Windows developers who know (and love) C++ be brought forward to use the .NET Framework? One answer is to extend C++, an option described later. Another approach, one that has proven more appealing for most C++ developers, is to create a new language based on the CLR but with a syntax derived from C++. This is exactly what Microsoft did in creating C#.

Note - C# is the natural language for .NET Framework developers who prefer a C-based syntax

C# looks familiar to anyone accustomed to programming in C++ or Java. Like those languages (and like the .NET version of VB), C# is object-oriented. Among the more interesting aspects of C# are the following:

  • Support for implementation inheritance, allowing a new child class to inherit code from one parent class, sometimes referred to as single inheritance
  • The ability for a child class to override one or more methods in its parent
  • Support for exception handling
  • Full multithreading (using the .NET Framework class library)
  • The ability to define interfaces directly in C#
  • Support for properties and events
  • Support for attributes, allowing features such as transaction support and Web services to be implemented by inserting keywords in source code
  • Garbage collection, freeing developers from the need to destroy explicitly objects they have created
  • Support for generic types, a concept similar to templates in C++
  • The ability to write code that directly accesses specific memory addresses, sometimes referred to as unsafe code

C++ or Java developers, who are accustomed to a C-like syntax, usually prefer C# for writing .NET Framework applications. VB6 developers, however, fond of their own style of syntax, often prefer the .NET version of VB, described next.

Visual Basic

From its humble beginnings at Dartmouth College in the 1960s, Basic has grown into one of the most widely used programming languages in the world today. This success was due largely to the popularity of Microsoft's Visual Basic. Yet it's likely that the original creators of Basic wouldn't recognize their child in its current form. The price of success has been adaptation-sometimes radical-to new requirements.

Note - Visual Basic is a very widely used language

VB.NET, the first CLR-based version of this language, was a huge step in Microsoft's evolution of VB. In addition to being enormously different from the original Basic language, it was a big leap from its immediate predecessor, VB6. The primary reason for this substantial change was that VB.NET was built entirely on the CLR. The version in Visual Studio 2005, officially known as Visual Basic 2005, brings a few changes to the original VB.NET. On the whole, however, this version of the language is very much like the VB.NET incarnation that preceded it.

Note - VB.NET was a big change from VB6

VB 2005 is much like C#, which shouldn't be surprising. Both are built on the CLR, and both expose the same core functionality. The biggest difference between C# and VB today is in syntax; functionally, the two languages are very similar. In fact, the list of interesting features in VB mirrors C#'s list:

  • Support for single implementation inheritance
  • Method overriding
  • Support for exception handling
  • Full multithreading (using the .NET Framework class library)
  • The ability to define interfaces directly in VB
  • Support for properties and events
  • Support for attributes
  • Garbage collection
  • Support for generic types

Note - Visual Basic 2005 provides almost exactly the same features as C#

Unlike its sibling C#, VB doesn't allow creating unsafe code. The 2005 version of VB does, however, provide an addition called the My namespace, making it easier to perform some common operations. Despite these differences, it's important to realize how similar VB and C# really are. The old division between C++ and VB6, two radically different languages targeting quite different developer populations, is gone. In its place is the sleek consistency of two closely related CLR-based languages: VB and C#.

C++

C++ presented a challenge to the creators of .NET. To be able to build .NET Framework applications, this popular language had to be modified to use the CLR. Yet some of the core semantics of C++, which allows things such as a child class inheriting directly from multiple parents (known as multiple inheritance), conflict with those of the CLR. And since C++ plays an important role as the dominant language for creating non-Framework-based Windows applications, modifying it to be purely Framework specific wasn't the right approach. What's the solution?

Note - The semantics of C++ differ from those defined by the CLR

Microsoft's answer is to support standard C++ as always, leaving this powerful language able to create efficient processor-specific binaries as before. To allow the use of C++ to create .NET Framework applications, Microsoft added a set of extensions to the language. The C++ dialect provided with Visual Studio 2005 that includes these extensions is called C++/CLI, and it provides access to all features of the CLR. Using C++/CLI, a developer can write C++ code that takes advantage of garbage collection, defines interfaces, uses attributes, and more.

Note - .NET Framework applications can be created using C++/CLI

Because not all applications will use the .NET Framework, Visual Studio still allows building traditional Windows applications in C++. Rather than using C++/CLI, a developer can write standard C++ code and then compile it into processor-specific binaries. Unlike VB and C#, C++ is not required to compile into MSIL.

Note - Standard C++ can also be used to create non-Framework-based applications

Other CLR-Based Languages

It's fair to say that the great majority of .NET development is done in C#, VB .NET, and C++. Still, Visual Studio 2005 also supports two other CLR-based languages that are worth mentioning:

  • JScript: JScript is based on ECMAScript, the current standard version of the language originally christened as JavaScript. As such, JScript provides developers with a loosely typed development environment, support for regular expressions, and other aspects typical of the JavaScript language family. Yet because it's based on the CLR, the .NET version of JScript also implements CLR-style classes, which can contain methods, implement interfaces, and more. While JScript can be used for creating essentially any .NET application, the language is most commonly applied in building ASP.NET pages.
  • J#: Microsoft's implementation of the Java programming language, J# provides Java's syntax on top of the CLR. Don't be confused-J# isn't intended for creating applications that run on a standard Java virtual machine. Microsoft also doesn't support many of the major Java libraries, such as Enterprise JavaBeans (EJB). Instead, J# provides a way to more easily transition Java developers and existing Java code to the world of .NET.

Note - Visual Studio 2005 also supports JScript .NET and J#

Microsoft also makes available a CLR-based version of the increasingly popular Python language. While it's not an integral part of Visual Studio, this implementation does allow access to the .NET Framework class library. It also demonstrates the ability of the CLR to support a broader range of programming languages, since Python is a dynamic language. Dynamic languages allow more change at runtime than more conventional approaches such as C# and VB, including things such as introducing new data types. To provide this kind of flexibility, the CLR-based version of Python relies on an interpreter rather than just a compiler.

Note - Microsoft provides Python for the CLR

Other vendors also provide tools and languages for creating .NET Framework applications. The most popular of these is probably Borland's Delphi. Widely admired as a language, Delphi has a hard-core user base around the world. The current version of this language is based on the CLR, and so it allows Delphi developers access to the services provided by the .NET Framework.

Note - Other vendors, such as Borland, provide CLR-based languages

Support for multiple programming languages is one of the most interesting things about the .NET Framework. Chapter 3 takes a closer look at C#, VB, and C++/CLI, the three most popular languages for building .NET Framework applications.

Domain Specific Languages

General purpose programming languages such as C# and VB are the mainstays of modern development. Yet the idea of model-driven development (MDD), where software creation depends at least in part on some kind of underlying abstract model, is a hot idea in the industry today. Visual Studio 2005 reflects this in its support for what are known as domain specific languages (DSLs). In general, a DSL might be expressed in text, graphically, or in some other way. For example, SQL can be thought of as a DSL designed for working with relational databases, while the graphical Orchestration Designer in Microsoft's BizTalk Server can be viewed as a DSL for defining some kinds of business logic. However it's defined, each DSL implements a set of higher-level abstractions for solving problems in a particular area.

Note - Domain specific languages can help in moving toward model-driven development

Visual Studio 2005 supports several graphical DSLs. Each is implemented in a specific graphical tool that focuses on a particular problem domain. For example, every version of Visual Studio 2005, except the Express Editions, includes a tool called the Class Designer. As Figure 1-9 illustrates, this tool provides a graphical DSL for creating and modifying classes and class hierarchies.

Figure 1-9 - The Visual Studio 2005 Class Designer provides a DSL for defining classes and class hierarchies.

Note - Visual Studio 2005 provides a Class Designer and other DSLs

The Class Designer allows a developer to create a new class by dragging and dropping an icon from the toolbox (shown on the left in Figure 1-9) onto the design surface. Properties, methods, and other aspects of the class can also be added graphically. This is more than just a tool for drawing pictures, however. The Class Designer actually generates code that reflects the defined classes. Changes to the diagram are reflected in the code, and changes made directly to the code are also reflected in the diagram. Especially for more complex scenarios, the ability to visualize classes and their relationships can be a big help in writing good code.

Note - The Class Designer DSL generates code from class diagrams

Perspective: Microsoft's Approach to Model-Driven Development - The dream of creating software through pictures has been with us for decades. Yet especially when they're expressed graphically, it's easy to look at DSLs and be reminded of the failure of computer-aided software engineering (CASE). Now that memories of the CASE debacle are fading, are DSLs just a repeat of this idea being foisted on a new generation?

Maybe-it's too soon to know. But increased developer productivity is a laudable goal, one that's worth taking some risks to achieve. The CASE technologies of the 1980s were far broader in scope than the relatively simple DSLs that are included in Visual Studio 2005. By avoiding the grand ambition of these earlier tools, Microsoft lessens the risk that its DSL efforts will meet the same fate. And the idea of creating at least some code graphically, especially code that targets a particular domain, is very attractive.

Microsoft is taking a small step in the DSL direction with Visual Studio 2005. The company's stated goal is to make developers perceive that modeling has value, not to move to a completely model-based development environment. Given this, starting small seems wise. Building this approach into one of the world's most widely used developer tools will certainly provide a solid test of the idea.

Microsoft isn't alone in promoting MDD, however. The Object Management Group (OMG), a multivendor consortium, has created a set of specifications for MDD. These specs define the OMG's Model Driven Architecture (MDA), and they have been endorsed by a number of vendors and user organizations. Yet Microsoft has essentially ignored MDA, choosing instead to go its own way.

Microsoft and the OMG have never had an especially cordial relationship. OMG's first creation was the Common Object Request Broker Architecture (CORBA), a direct competitor to Microsoft's DCOM. Like CORBA, MDA takes an explicitly cross-platform, vendor-neutral approach. In fact, a primary MDA goal is to create models that can be implemented equally well using different technologies, such as .NET and Java. Given its diverse membership, it's not surprising that OMG starts from this perspective.

It's also not surprising that Microsoft doesn't see the value in this. Visual Studio is a tool for building Windows applications, so why complicate things by implementing an MDD approach that strives to be cross-platform? And if you're Microsoft, what value is there in working with a committee of others to define what will ultimately be a Windows technology? Unsurprisingly, Microsoft continues to believe that for Windows programming interfaces, it should be in the driver's seat. Also, like the products of many standards committees, MDA is abstract and complex. (A friend of mine who's spent a considerable amount of time in the MDA world has described it as "where the rubber meets the sky.") While ignoring MDA might make Visual Studio integration with non-Microsoft tools more challenging, Microsoft's resistance to a technology created by its competitors is entirely in line with its usual perspective on the world. And as always, competition among different approaches is likely to make this technology improve more quickly.

Working in Groups: Visual Studio Team System

Like any development tool, the goal of Visual Studio is to help developers be more effective. Yet the majority of software today, and virtually all enterprise software, isn't created by a single individual. Software development has become a team sport, complete with specialized positions for various players.

Note - Most software today is created by groups, not individuals

Note -Visual Studio 2005 Team System includes several different tools

Visual Studio 2005 Team System recognizes this fact. As Figure 1-10 shows, this product suite includes components that target each member of a modern development group. Those components include:

Figure 1-10 - Visual Studio Team System provides different tools for different roles in a development team.
  • Visual Studio 2005 Team Edition for Software Architects: provides a group of tools known collectively as the Distributed System Designers. Each provides a DSL aimed at architects designing various aspects of service-oriented applications. These tools include Application Designer for defining applications and how they communicate, System Designer for defining how those applications fit together for deployment, Logical Datacenter Designer for defining the structure of machines in a data center, and Deployment Designer for defining how the set of applications that comprise a particular system is deployed in a specific data center. This product also includes a version of Visio that allows creating Unified Modeling Language (UML) diagrams, entity/relationship (ER) diagrams, and other architecture-oriented pictures.
  • Visual Studio 2005 Team Edition for Software Developers: includes tools that are useful for people who actually write code. These tools support static code analysis, exposing problems such as using an uninitialized variable, dynamic code analysis, allowing code to be profiled for performance improvements, and more.
  • Visual Studio 2005 Team Edition for Software Testers: offers tools focused on the tasks performed by people who test code, such as tools for creating and running unit tests and for load testing.
  • Visual Studio 2005 Team Foundation Server: provides the common platform for the other Team System components. Implemented as a standalone server application, Team Foundation Server keeps track of team projects, maintains a work items database, supports version control of a project's source code, and provides other common services that a software development team needs. The projects maintained by a particular Team Foundation server can also be examined using a client called the Team Explorer. Status reports and other information are available through a Team System portal built on Windows SharePoint Services.

Except for Team Foundation Server, all of the components of Visual Studio 2005 Team System are built on Visual Studio 2005 Professional Edition, which means that each one includes a complete development environment. While many .NET developers are happy with just a standalone version of Visual Studio, larger teams or those working on more complex projects can sometimes make good use of the extra tools provided by Visual Studio 2005 Team Edition.

Conclusion

Since its original release in 2002, the .NET Framework has become the foundation for a majority of new Windows applications. Judging from the evidence so far, it has surely been a success for Microsoft and for its customers. While the move to the .NET environment forces developers to climb a long learning curve, the benefits appear to be worth the effort. For the people who use it, this technology qualifies as one more step toward the ultimate goal: producing the best possible software in the least amount of time.

The Pain of Change - During a .NET seminar I gave in Moscow a few years ago, one of the participants raised his hand with a concerned expression. "I'm an experienced Windows DNA developer," he said. "If I learn this .NET stuff, can you promise me that this is the last new Microsoft technology I'll ever have to learn?"

I couldn't, of course. What I could promise him was that he was in the wrong profession. Even if my worried questioner sticks with the Microsoft platform for the rest of his career, it's all but certain that new technologies will appear that he'll need to understand. As long as the hardware we depend on keeps getting better, and as long as creative people work in this field, new software technologies will continue to appear.

Fortunately, changes as large as .NET aren't common. Bringing out new languages, a large new library, and significant revisions to other core technologies all at once, as Microsoft did with .NET, was almost too much to swallow. Yet bringing out those same changes piecemeal would likely have been worse, if only because the integration among them would certainly have suffered. To make progress, vendors are sometimes forced to make their customers swallow a large amount of change all at once.

The .NET environment is now a standard part of the software world. Still, don't think it's the last word in software technology-it's not. If you don't like change, get out of the software business.

Footnotes

1 Pronounced "C sharp," as in the musical note
  2 "SOAP" was originally an acronym for "Simple Object
    Protocol." Today, the standards group that owns this
    technology has decided that SOAP no longer stands for
    anything-it's just a name.