An interesting question is whether there exists a diff algorithm
which *minimizes* data movement across the network. Assuming similar
lengths, you can halve it by running the diff at one of the hosts
but can one do better if the two files are fairly similar? Is this
even a theoretical possibility? I don't see links to any such
algorithm on wikipedia's diff page but I figured there might be
someone on TUHS who may have speculated or know about this!

Bakul