A discussion on Mastodon about the file formats of backups led me to
think it through again:
First things first: I have two sets of hard disks for backups, one for
each laptop. We keep one set at the office at all times. We make
occasional backups on one set, and take it to the office, and we bring
back the other set. This protects us against theft and fire.
I started using Borg Backup last year.
I don’t care about the file format used per se. I still feel that a
backup solution should be very robust and that robustness should be
more important than deduplication and all that, and that’s why I
really liked my old rsync based backup. But setting it all up to have
multiple linked snapshots would have been hard, too.
My measure of robustness is perhaps colored by the occurrence of bad
sectors in the old days of floppy disks. Thus, a format is robust if I
can still read the rest using a tool. That disqualifies compressed
data, encrypted data, or anything else that results in a dependency on
bytes in the damaged sectors (virtual file systems and the like).
Basically, anything but plain text. 🧐
In my situation, I decided I valued encryption more than robustness
because I am keeping external disks outside our home. Given that this
already results in an opaque byte soup, I decided I might as well use
any other format (Borg Backup, whatever). That’s for making backups of
my laptops to keep them outside our home.
On the other hand, when making backups of my servers to my laptop, the
situation is different: encryption is not essential, and plaintext is
robust, and that’s why I use rsync.
Anyway. The most important part is to make backups. Any kind of backup
will do. And check whether you know how to recover files from the
backup!
