Cutting Edge: DataSets vs. Collections -- MSDN Magazine, August 2005: "Typed DataSet—Good and Bad
Aware of the logical limitations of the DataSet object, Microsoft also introduced the typed DataSet—a class that derives from DataSet and inherits all the members of a DataSet. In addition, a typed DataSet provides strongly typed members to access tables and columns by name, instead of using generic collection-based methods. This is beneficial for at least two reasons. First, it improves the overall readability of the code and provides significant help from the Visual Studio 2005 IDE through IntelliSense® and automatic code completion. Second, typed DataSets let you distinguish one table from the next using different objects to render each. Table Employees, for example, will be a different object from table Customers. In this way, type mismatch errors are caught at compile time rather than during execution.
Typed DataSets are still data containers, they're just a bit less generic and have a little more information about the data they contain. They can still hold any data, but you get some specialized members to speed up any work you need to do on a few particular types of data.
Typed DataSets include the same serialization algorithm as DataSets, but because they're derived classes it's easier to extend them further with manually written code to improve the serialization mechanism. For example, your typed DataSet can reimplement the ISerializable interface to reduce or compress the amount of data being moved if you find out that in the particular context in which it operates the XML-based serialization algorithm is too heavy. Speaking of this, let me add a brief remark. As Figure 1 shows, the XML algorithm performance is not so bad when you're moving only a few hundred rows. Sure, each system has its own size and data, but if you realize you're moving thousands of rows across the layers, spend some time making sure you're doing it right. It m"