No real answer but some thoughts:
To do an in-place update of some remote copy of your file, you have to first find out which parts of the file have changed or have been added since the last copy. One could maintain a list of checksums for chunks of n megabytes. Without such a list, you have to read the remote copy and the local copy to find the changes.
Drawback of an in-place update would be the inherent risk of data corruption. If the update fails or is aborted, the remote copy is left in an inconsistent state.
The choice of a good replication strategy depends on several parameters:
- Is it feasible to read the remote copy efficiently?
What is the speed to do this?
- Are changes scattered across the whole file or more or less local?
- What is the probability that changes in a chunk of data go unnoticed due to clashes in
The checksum algorithm has to be chosen in such a way that this is sufficiently unlikely.
I think you have a good and valid point in your idea. It would be very practical - for example - to save big Outlook .pst files on a memory stick. However, I am not aware of any tool or method which solves all the obstacles mentioned above.