Typed versus Untyped Datasets
Datasets can be typed or untyped. A typed dataset is a dataset that is first derived from the base DataSet class and then uses information in an XML Schema file (an .xsd file) to generate a new class. Information from the schema (tables, columns, and so on) is generated and compiled into this new dataset class as a set of first-class objects and properties.
Note For more information about dataset schemas, see XML Schemas and Data.
Because a typed DataSet class inherits from the base DataSet class, the typed class assumes all of the functionality of the DataSet class and can be used with methods that take an instance of a DataSet class as a parameter
An untyped dataset, in contrast, has no corresponding built-in schema. As in a typed dataset, an untyped dataset contains tables, columns, and so on — but those are exposed only as collections. (However, after manually creating the tables and other data elements in an untyped dataset, you can export the dataset's structure as a schema using the dataset's WriteXmlSchema method.)
You can use either type of dataset in your applications. However, Visual Studio has more tool support for typed datasets, and they make programming with the dataset easier and less error-prone.
Contrasting Data Access in Typed and Untyped Datasets
The class for a typed dataset has an object model in which its tables and columns become first-class objects in the object model. For example, if you are working with a typed dataset, you can reference a column using code such as the following:
// This accesses the CustomerID column in the first row of
// the Customers table.
s = dsCustomersOrders1.Customers.CustomerID;
In contrast, if you are working with an untyped dataset, the equivalent code is:
string s = (string) dsCustomersOrders1.Tables["Customers"].Rows["CustomerID"];
Typed access is not only easier to read, but is fully supported by IntelliSense in the Visual Studio Code Editor. In addition to being easier to work with, the syntax for the typed dataset provides type checking at compile time, greatly reducing the possibility of errors in assigning values to dataset members. Access to tables and columns in a typed dataset is also slightly faster at run time because access is determined at compile time, not through collections at run time.
Even though typed datasets have many advantages, there are a variety of circumstances under which an untyped dataset is useful. The most obvious scenario is that no schema is available for the dataset. This might occur, for example, if your application is interacting with a component that returns a dataset, but you do not know in advance what its structure is. Similarly, there are times when you are working with data that does not have a static, predictable structure; in that case, it is impractical to use a typed dataset, because you would have to regenerate the typed dataset class with each change in the data structure.
Hope this will help you.