Re: [Duplicity-talk] Suggested increment chain

[Kenneth Loafman]

Marcus Williams wrote:
> On 13/02/2009 14:37, Kenneth Loafman wrote:
>> As to frequency, I would suggest a full every week and incrementals
>> every day.  This means that if a backup should be bad, you won't be out
>> more than a week or so from the last recoverable backup.
>>
>> There is no "right" answer, so its all tunable to the usage.
> 
> Is there anyway to create a diff with duplicity? Or get an incremental
> that ignores all current incrementals and performs an incremental purely
> against the last full backup (so basically is a diff)?

If you have access to the target directory, you could move the previous
incremental backup files out of the way, perform an incremental, and get
the duplicity equivalent of a differential.  However, and this is very
important, the resulting incremental backup would not be the same as a
differential backup.  A differential backup creates a complete copy of
the changed file.  A duplicity incremental backup creates a difference
set of changes to the original file.

This would be the equivalent of coalescing the incremental sets
together, which is something I've thought of tackling, but have not
since I think it would lead to bad backup behavior in that full backups
would not be done on a regular basis and too much reliance would be put
on ancient backups and long chains of incremental/differential backups.
 That's a recipe for disaster.

Plus, I've had experience in legal discovery methods (primary author of
the first Synthetix (tm) and Syndex (tm) packages).  Believe me when I
tell you that you do want to destroy old backups unless you have strong
legal reasons to hold on to them.  Keeping old backups around is
courting legal disaster, even if they are encrypted.

...Ken

signature.asc
Description: OpenPGP digital signature