A suitable find command:

• with -printf (GNU find has it)
• or with -print0 (BSD and GNU find's have it)
• or a traditional (SysV, early BSD) find with -print combined with a "tr" command that can translate newlines to nulls. This breaks on filenames containing newlines of course.

Compression

• Backshift will prefer to compress chunks from greatest preference to least:
  1. The CPython 3.3 "lzma" module. This gives xz-compatible compression.
  2. liblzma (likely in a -dev package) via ctypes. This gives xz-compatible compression.
  3. An xz binary via the subprocess module.
  4. The Python bz2 (bzip2) module.
• xz/lzma give a bit better compression ratios than bzip2.
• Notes:
  • The Debian python-lzma package is not the same as the new CPython 3.3 lzma module; if you have python-lzma installed, you'll probably need to uninstall it.
  • You can tell what kind of compression is being used (in 1.21 and up) with:
    • backshift --compression-means

Filesystems - for remote saves

• sshfs
• CIFS
• NFS
• ...but see the backshift notes for your OS, if any.
• Others are probably possible, but have not been tested. Perhaps Lustre, GFS, PVFS2, OCFS2...?

Time synchronization

• It is recommended that you use NTP or similar technology for keeping your clocks synchronized. Backshift relies heavily on correct timestamps as files are re-marked in-use.
• You may also want to use something like check-clocks, to compare your clock times prior to each backup.

For building and/or testing:

• Some form of "make"
• You may want SVN (Subversion)
• You want an m4 binary with which to create the pure python and/or cython rolling checksum modules