[Duplicity-talk] Question about the TAR replacement format

[Elvis Stansvik]

Hi all,

I got interested in the archive format described at [1].

Is there still an interest in developing this format further and eventually replacing the use of TAR in Duplicity?

I have no particular experience with archive formats, but I have a couple of questions (some of which may be dumb):

* Is the inner index file (the one containing repeated metadata) supposed to come after the inner data file? I've attached a full diagram of the format to show the order of things. Is it correct?

* Why the division into such small (<= 65k) blocks at the outer level? Why not let each block at the outer level contain an entire inner data/index pair?

* The notes says: "Archive footer: Must contain all information in the header, in case the header is lost.". When would the header get lost?

* Would it not be useful to have some blocks compressed and some not (those corresponding to already compressed files)?

I saw in the archives that a few years ago someone suggested maybe using XAR [2], but got no response. Could that be an alternative or is there something missing that Duplicity would need? The XAR format seems quite badly documented though..

Best regards,

Elvis Stansvik

[1] http://duplicity.nongnu.org/new_format.html

[2] https://code.google.com/p/xar/

format.png
Description: PNG image