From a high level, the method control layer is responsible for orchestrating network operations and managing the network methods. This includes several responsibilities, including address resolution, method multiplexing, and providing a stable BMI user interface. It also provides a library of support functions that may be useful to method implementors.
One of the most important tasks of the method control layer is the multiplexing of network methods. When an operation is posted by the user, it is up to the method control to decide which method will service the operation. Likewise, when the user tests for completion, the method control must test the appropriate methods for the operations of interest.
The method control layer provides the BMI user interface. This is the API used by applications that communicate using BMI. The BMI interface functions are converted into the appropriate low level method requests that are needed to complete operations.
Address resolution is the final major responsibility of the method control. The method control manages the BMI level addresses and makes sure that the name space is consistent to the user, regardless of which methods are in use. It does so by maintaining an internal reference list for addresses. Each network address has a unique reference that provides mappings between BMI user level addresses, the string representation of addresses, and the method specific representation of addresses. The BMI user level addresses are handles for network hosts that the application uses when calling BMI functions. The string representation is the ASCII host name of the hosts before they are resolved by BMI (as read from a ``hosts'' file, for example). Finally, the method address is the representation that that methods use for identifying hosts, which may contain information specific to that particular protocol. Note that method addresses are never, under any circumstances, exposed to the application. They are reserved for internal BMI use only.