Pipelining

The above sections describe the process for constructing a copy of one file on a remote system. If we have a several files to copy, we can gain a considerable latency advantage by pipelining the process.

This involves $\beta$ initiating two independent processes. One of the processes generates and sends the checksums to $\alpha$ while the other receives the difference information from $\alpha$ and reconstructs the files.

If the communications link is buffered then these two processes can proceed independently and the link should be kept fully utilised in both directions for most of the time.