SparseFiles

From RdiffBackupWiki

(text deleted)..e files nicely would be lovely, but there's some evidence it is not currently possible to do it in a portable fashion.

-- Andrew Bressen <bressen@savannah.nongnu.org>

Rather than worrying about copying a file's sparseness you could just make all output files sparse by default - so when writing, write in blocks and deal with a block of zeroes as a nop followed by a seek to the next set of non-zero data.

-- Nigel Metheringham <nigel.metheringham@pobox.com>

Implementing sparse files that way may save some HD space, but I think in order for it to be worth it, sparse files need to be detected on the source site. For instance, I had some bug on my machine and one of the log files became 1+TB long, almost all of it a "hole". So even if rdiff-backup wrote a sparse file on the destination, it would still take absurdly long to process the file, and to transfer it.

Also even if rdiff-backup could detect spare files, it would take a modification of the rsync algorithm in order to patch/diff spare files efficiently. As it is now they would just process buffer after buffer full of 0's.

Anyway, Nigel's suggestion is a good one and relatively easy to implement.

I guess it depends on how/why sparse files are used, and what the consequences of not supporting them are.

-- BenEscoto