It may be assumed that a DLA has rather extensive knowledge about the domain where it was created. For example, the topology of the network may be known to the application injecting the DLA and hence programmed into the DLA. Such domain knowledge can no longer be assumed when DLAs migrate through the larger network in order to implement global policies. A DLA may know that certain endpoints are connected to a particular MCA domain, but it generally has no knowledge of the switches and interconnections on the paths between these endpoints. This makes it impossible to exercise low-level control over the resources in this domain (using netlets), unless we provide the netlet with operations to learn about its new environment.
For this purpose, the tertiary interface contains a number of operations to allow DLAs to acquire knowledge about the new domain. One of these operations returns to the DLA (upon request) a detailed description of a path between two endpoints. Using this operation, a DLA can learn about the part of the MCA domain that is of interest to the DLA. The DLA can now start creating netlets consisting of exactly these paths, allowing it the low-level control that it may require to implement application specific policies in the new domain.