Proposed new format - why not dar?

Asked by Chris on 2019-03-03

I see that authors discuss a new proposed archive format on the project's website: http://duplicity.nongnu.org/new_format.html

I was wondering what is the author's opinion about dar: http://dar.linux.free.fr/

It is very featured archive format providing pretty much all required features: index, in-archive compression and encryption, index extraction (you can download only the index part for local filesystem diff-ing), C library and command line tool.

Question information

Language:
English Edit question
Status:
Solved
For:
Duplicity Edit question
Assignee:
No assignee Edit question
Solved by:
Aaron Whitehouse
Solved:
2019-03-04
Last query:
2019-03-04
Last reply:
2019-03-03

Please see the mailing list threads starting here:
https://lists.nongnu.org/archive/html/duplicity-talk/2015-08/msg00064.html
https://lists.nongnu.org/archive/html/duplicity-talk/2015-09/msg00006.html

Since then, it looks as though the rsync-like increment has been implemented:
https://sourceforge.net/p/dar/feature-requests/122/

There was also a reference to dar here:
https://answers.launchpad.net/duplicity/+question/150909

The short answer is that all the main contributors seemed in favour of the format. We should probably remove that page, as it does keep coming up.

Separate to how good the format is, there is the question of using the format in duplicity. While I am generally in favour of trying, I have a few hesitations:
* The main problem here is that, while Dar seems to be regularly updated, the python bindings are not.
* While the format seems good from a quick look, I saw at least one project move from Dar to duplicity because the output files could be managed with 'normal' GNU/Linux tools. Dar files are clearly less common than tar files.
* If duplicity fully embraces Dar, duplicity would essentially become a wrapper adding things like the backends. In many ways that would be good, but it does mean that we would lose the ability to fix bugs etc in this project (and python-only contributors would be unlikely to be able to work on Dar itself).

Chris (ezaquarii) said : #3

Thanks Aaron Whitehouse, that solved my question.