Transports and Semantics

The RPC protocol is independent of transport protocols. That is, RPC disregards how a message is passed from one process to another. The protocol handles only specification and interpretation of messages.

RPC does not attempt to ensure transport reliability. Therefore, you must supply the application with information about the type of transport protocol used under RPC. If you tell the RPC service that it is running on top of a reliable transport such as TCP, most of the work is already done for the service. On the other hand, if RPC is running on top of an unreliable transport such as UDP, the service must devise its own retransmission and time-out policy. RPC does not provide this service.

Because of transport independence, the RPC protocol does not attach specific semantics to the remote procedures or their execution. Semantics can be inferred from, but should be explicitly specified by, the underlying transport protocol. For example, suppose RPC is running on top of an unreliable transport. If an application retransmits RPC messages after short timeouts receiving no reply, it can infer only that the procedure was executed zero or more times. If the application does receive a reply, it can infer that the procedure was executed at least once.

A server might choose to remember previously granted requests from a client and not regrant them to ensure some degree of execute-at-most-once semantics. A server can do this by using the transaction ID that is packaged with every RPC request. The main use of this transaction ID is by the RPC client for matching replies to requests. However, a client application can choose to reuse its previous transaction ID when retransmitting a request. The server application, checking this fact, can choose to remember this ID after granting a request and not regrant requests with the same ID. The server is not allowed to examine this ID in any other way except as a test for equality.

On the other hand, if using a reliable transport such as TCP, the application can infer from a reply message that the procedure was executed exactly once. If the application receives no reply message, it cannot assume the remote procedure was not executed. Note that even if a connection-oriented protocol like TCP is used, an application still needs timeouts and reconnection to handle server crashes.