Mapped Drive

[Matt McCutchen]

On 10/1/07, Matt McCutchen <hashproduct+rsync at gmail.com> wrote:
> If the remote filesystem supports efficient copying of a range of data
> from one offset to another, then #2 is moot and a smart client can do
> both pushes and pulls efficiently using your scheme and zsync's
> "reverse" delta-transfer algorithm, respectively.

An alternative approach is to store the blocks of each file as
separate files on the remote filesystem.  For example, a file named
"foo" of 1 MB might become files named "foo.1", ..., "foo.16" of 64 KB
each.  This works on even a completely dumb filesystem; a block of
data can be "moved" from one offset to another by renaming the file
holding it.  However, this approach requires all clients to have
special software to interpret the split files, and it may have
unacceptable overhead depending on the remote filesystem
implementation.  At that point, it may make more sense to just run an
rsync daemon on the remote machine.

Matt