In this discussion, I attempt to isolate data types from other C++ types. The C++ language contains several built-in primitive data types that include various sizes of integer and floating point objects. These types are fundamental to the language and are the foundation upon which every C++ program is built. However, these built-in types are constrained and bound to the address space occupied by the C++ program. This allows the language to exploit addressing and the native capabilities of the processor. Addresses are treated very much like data from the machine point-of-view.

When information is moved from one address space to another, it becomes necessary to carefully distinguish address/location information from value information.

More generally, data can be copied between address spaces. When this is done, absolute location/address information becomes useless in the new address space context. As a result,
 

• Addresses are meta data. An address describes the location of a data item.
• Type information is meta data

I believe there is a certain class of C++  constructs that I will call a Data Manipulator. Data manipulators have the following characteristics:

• Adds no memory overhead to the raw data type. In general, this means that no meta data information can be stored with the object. Examples of this excluded information include vtables, base addresses, and type information.
• Must declare no constructors.

The impact of these constraints in terms of C++ are:

• Data manipulator cannot use polymorphic facilities including:
• Inheriting virtual operations either directly or indirectly.
• Data manipulator base classes must declare the memory space for the data completely.