Before examining the specifics of the .NET universe, it's helpful to consider some of the issues that motivated the genesis of Microsoft's current platform. To get in the proper mind-set, let's begin this chapter with a brief and painless history lesson to remember our roots and understand the limitations of the previous state of affairs (after all, admitting you have a problem is the first step toward finding a solution). After completing this quick tour of life as we knew it, we turn our attention to the numerous benefits provided by C# 2010 and the .NET 4.0 platform.
Traditionally speaking, developing software for the Windows family of operating systems involved using the C programming language in conjunction with the Windows application programming interface (API). While it is true that numerous applications have been successfully created using this timehonored approach, few of us would disagree that building applications using the raw API is a complex undertaking.
The first obvious problem is that C is a very terse language. C developers are forced to contend with manual memory management, ugly pointer arithmetic, and ugly syntactical constructs. Furthermore, given that C is a structured language, it lacks the benefits provided by the object-oriented approach (can anyone say spaghetti code?). When you combine the thousands of global functions and data types defined by the Windows API to an already formidable language, it is little wonder that there are so many buggy applications floating around today.
One vast improvement over raw C/API development is the use of the C++ programming language. In many ways, C++ can be thought of as an object-oriented layer on top of C. Thus, even though C++ programmers benefit from the famed pillars of OOP (encapsulation, inheritance, and polymorphism), they are still at the mercy of the painful aspects of the C language (e.g., manual memory management, ugly pointer arithmetic, and ugly syntactical constructs).
Despite its complexity, many C++ frameworks exist today. For example, the Microsoft Foundation Classes (MFC) provides the developer with a set of C++ classes that facilitate the construction of Windows applications. The main role of MFC is to wrap a sane subset of the underlying Windows API behind a number of classes, magic macros, and numerous code-generation tools (aka wizards). Regardless of the helpful assistance offered by the MFC framework (as well as many other C++-based windowing toolkits), the fact of the matter is that C++ programming remains a difficult and error-prone experience, given its historical roots in C.
Due to a heartfelt desire to enjoy a simpler lifestyle, many programmers shifted away from the world of C(++)-based frameworks to kinder, gentler languages such as Visual Basic 6.0 (VB6). VB6 became popular due to its ability to build complex user interfaces, code libraries (e.g., COM servers), and database access logic with minimal fuss and bother. Much more than MFC, VB6 hid the complexities of the raw Windows API from view using a number of integrated code wizards, intrinsic VB data types, classes, and VB-specific functions.
The major downfall of VB6 (which has been rectified given the advent of the .NET platform) is that it is not a fully object-oriented language; rather, it is object based. For example, VB6 does not allow the programmer to establish is-a relationships between classes (i.e., no classical inheritance) and has no intrinsic support for parameterized object construction. Moreover, VB6 doesn't provide the ability to build multithreaded applications unless you are willing to drop down to low-level API calls (which is complex at best and dangerous at worst).
Note The Visual Basic language used within the .NET platform (which is often referred to as VB.NET), has very little relationship to VB6. For example, modern day VB supports operator overloading, classical inheritance, type constructors and generics.
Enter Java. Java is an object-oriented programming (OOP) language with its syntactic roots in C++. As many of you are aware, Java's strengths are far greater than its support for platform independence. As a language, Java cleans up many unsavory syntactical aspects of C++. As a platform, Java provides programmers with a large number of predefined packages that contain various type definitions. Using these types, Java programmers are able to build "100% Pure Java" applications complete with database connectivity, messaging support, web-enabled front ends, and a rich desktop user interface (among other services).
Although Java is a very elegant language, one potential problem is that using Java typically means that you must use Java front to back during the development cycle. In effect, Java offers little hope of language integration, as this goes against the grain of Java's primary goal—a single programming language for every need. In reality, however, there are millions of lines of existing code out there in the world that would ideally like to commingle with newer Java code. Sadly, Java makes this task problematic. While Java does provide a limited ability to access non-Java APIs, there is little support for true cross-language integration.
The Component Object Model (COM) was Microsoft's previous application development framework, which first appeared on the programming landscape circa 1991 (or 1993, if you regard COM's introduction with the birth of OLE 1.0). COM is an architecture that says, in effect, "If you build your types in accordance with the rules of COM, you end up with a block of reusable binary code." These binary blobs of COM code are often called "COM servers".
One benefit of a binary COM server is that it can be accessed in a language-independent manner. Thus, C++ programmers can build COM classes that can be used by VB6; Delphi programmers can use COM classes built using C, and so forth. However, as you may be aware, COM's language independence is somewhat limited. For example, there is no way to derive a new COM class using an existing COM class (as COM has no support for classical inheritance). Rather, are limited to reuse via the has-a relationship.
Another benefit of COM is its location-transparent nature. Using constructs such as the system registry, application identifiers (AppIDs), stubs, proxies, and the COM runtime environment, programmers can avoid the need to work with raw sockets, RPC calls, and other low-level details when building a distributed application. For example, consider the following VB6 COM client code:
' The MyCOMClass class could have be written in
' any COM-aware language, and may be located anywhere
' on the network (including your local machine).
Dim obj as MyCOMClass
Set obj = New MyCOMClass ' Location resolved using AppID.
obj.DoSomeWork
Although COM can be considered a very successful object model, it is extremely complex under the hood (at least until you have spent many months exploring its plumbing—especially if you happen to be a C++ programmer). To help simplify the development of COM binaries, programmers can make use of numerous COM-aware frameworks. For example, the Active Template Library (ATL) provides a set of C++ classes, templates, and macros to ease the creation of COM servers.
Many other languages also hide a good part of the COM infrastructure from view. However, language support alone is not enough to hide the complexity of COM. Even when you choose a relatively simple COM-aware language such as VB6, you are still forced to contend with fragile registration entries and numerous deployment-related issues (collectively, and somewhat comically, termed DLL hell).
Although COM certainly facilitates the construction of software applications using a variety of different programming languages, the language-independent nature of COM was not as straightforward as one would hope.
Some of the complexity is due to the simple fact that applications that are woven together using diverse languages are completely unrelated from a syntactic point of view. For example, JScript has a syntax much like C, while VBScript is a subset of VB6. The COM servers that are created to run under the COM+ runtime (a Windows OS feature which provides common services to custom code libraries [transactions, object lifetime, security, etc]) have an entirely different look and feel from the web-centric ASP pages that invoke them. The result was a rather confused mishmash of technologies.
Furthermore, and perhaps more important, each language and/or technology has its own type system (that may look nothing like another’s type system). Beyond the fact that each API ships with its own collection of prefabricated code, even basic data types cannot always be treated identically. A CComBSTR in ATL is not quite the same as a String in VB6, both of which have nothing to do with a char* in C.
Given that each language has its own unique type system, COM programmers typically needed to be very careful with building public methods on public COM classes. For example, if a C++ developer needed to create a method that returned an array of integers to a VB6 application, they would be up to their eyeballs in complex COM API calls to construct a SAFEARRAY structure, which could easily require dozens of line of code. In the COM world, the SAFEARRAY data type is the only way to build an array that all COM frameworks understand. If the C++ developer simply returned a native C++ array, VB6 would have no clue what to do with it.
Similar complexities could be found when building methods that manipulate simple string data, references to other COM objects, or even a trivial BOOLEAN value! To put it politely, COM programming is a very asymmetrical discipline.