magnust, barring external factors (such as pre-Duplicati encryption) since Duplicati splits files up into smaller block sizes (check out the --blocksize parameter, not to be confused with --dblock-size) I believe what you should see is:
- Only changes within a block are resent (so in a 100M file if 1byte is changed and your block-size is 100k, then a single 100k block will be sent. However, if 1k is changed the actually re-send amount could be anywhere between a single 100k file and 1,024 of them depending on how closely the 1k of changes aligns with individual blocks. Most likely you wouldn’t see more than 2 or 3 individual blocks being re-sent)
- Yes, versioning works the way you describe except that the backup will increase in size not by the size of the changes made, by but by the size of the blocks necessary to “contain” the changes made (see answer #1).
- Because of the blocks, moving / renaming files will not cause the file (blocks) to be re-uploaded, however there will still be some small increase in backup size due to recording the “versions” of file names / locations.
Oh, and I started this with “barring external factors” because if there’s some other process running on the machine that might, say, encrypt or compress files before Duplicati processes them then it’s POSSIBLE that a small 1k change could cause the entire file to “look” different (encryption often does this). Note that I doubt many (any?) people will really fall into this scenario but I THINK it’s possible (please correct me if I’m wrong) so thought I’d mention it.
If you really want to understand what will happen in these scenarios, check out this page.