Re: rdiff-backup: Restoring a file from its increments

On 12/20/22 7:46 AM, Dieter Heußner via Any discussion of rdiff-backup wrote:
> Hello,
> 
> could somebody shed some light to the question of restoring files backed up by rdiff-
> backup, please?
> 
> Given an original file, say file A. It will be backed up with rdiff-backup, and also
> its increments (assuming that there are many changes to file A).
> 
> When restoring a given version of that file,
> a) does rdiff-backup construct the desired version step-wise by applying ALL
> increments of that file (in chronological order beginning with the original version), OR
> b)  does rdiff-backup start with the original file (file A) and ONLY applies the most recent
> increment which relates immediately prior or precisely to the timestamp of the
> desired version to be restored?
> 
> In my understanding, if a) was true, deleting increments (say older than n
> weeks) would not always result in the desired version because of probably missing
> increments. If b) was true, the goal can be achieved. However, as a newbie to rdiff-backup
> I am not sure at all.

rdiff-backup always starts with the _most_ _recent_ version of the file (the one stored in the mirror). If what you want is an earlier version, the stored reverse-diffs are applied in reverse chronological sequence (most recent diff first) until reaching the date you want. You can freely delete (using "rdiff-backup remove-increments --older-than ...", and _not_ by manually deleting files from the repository yourself) anything older than the history you want to preserve.

-- 
Bob Nichols     "NOSPAM" is really part of my email address.
                 Do NOT delete it.