Re: [Dar-support] Backupstrategy for optical disks

[<bun...@di...>]

Dear Denis,
dear list,


> Le 17/01/2017 à 20:47, bun...@di... a écrit :
> > 
> [...]
> > 
> > Are dar-files more robust as zip-files?
> 
> Answering this question correctly is not that easy. So the following
> test is to be taken with extreme caution: First thing, a single test
> is not representative, at most it may be taken as an indication not as
> a proof. Second thing, we can tell there is a conflict of interest as
> I am strongly linked with one of the two parties (dar).
> 
> For my onw curiosity, what I did is creating a dar and a zip archives
> of the same data...


thank you for your test. I've learned a lot! I used the time since your test (in january (!)) to read and learn a lot about backup and archives.

Having 140GB of images to backup I decided to use 7zip as splitted (each 70 GB) and encrypted archive (without any compression). Using dvdisaster allows me to add 30% of recovery data for each of the two 100GB-Blueray-MDisks.


> What to conclude with that, now?
> 
> As I said, generalizing from a single test is not a good idea. The
> only point I see that can be generalized is the flexibility dar brings
> in term of compression algorithm choice and here we see the gap in
> term of compression ratio. The other point that shows is the internal
> redundancy of dar archives for its metadata. Another thing that I can
> mention is the existence of the lax-mode that may help extracting some
> data from a really corrupted dar archive. As I do not know zip
> internals I cannot tell what's great in it so comparison is not
> balanced, take that into account reading this conclusion.
> 
> Last, care should be taken because the randomly chosen byte took place
> in the file's data for the zip archive while it took place in the
> sequential metadata in dar archive. If the corruption for dar had been
> in the data of a file, dar would not have acted differently from zip,
> may it be using direct or sequential reading mode: dar would have
> reported the file as corrupted to the user and would have restored any
> other files as well, but this should also be tested not just pretended
> as I do here.
> 

Again thank you for your tests. I'll take dar again into account if I have a different scenario. 

Regards!