I hate having to deal with back-ups, it's time-consuming. Things can
go wrong with them too, like what my web host did: Backed-up and
restored my site's files without telling me (they were probably doing
some maintenance on their gear the first time, the second time they
were flailing around in the dark after they'd destroyed their perl
installation). Every file had their permissions fouled up. Twice,
now, I've had to un-munge about 1500 files.
T'is a common tale but true.
I once spent a summer rescuing plain text data files from a backup of CDC Cyber
onto Solaris. The files had records out of order or missing and blocks of duplicated
records. Over the years, there had been changes to the data format.
The original files were used to produce printed data reports, then later scanned
with an automated system that sometimes messed up a page without anyone
checking. In some cases OCR on short sections of the data reports was able
to replace missing records.
I did the work using command-line tools in Apple OSX. Rather than manually
editing the files I was able to write shell scripts to remove duplicate records, sort
them correctly, and adjust to a common format. It was a big exercise in POSIX
shell and utilities text processing.