Cedar Backup began life in late 2000 as a set of Perl scripts called kbackup. These scripts met an immediate need (which was to back up skyjammer.com and some personal machines) but proved to be unstable, overly verbose and rather difficult to maintain.

In early 2002, work began on a rewrite of kbackup. The goal was to address many of the shortcomings of the original application, as well as to clean up the code and make it available to the general public. While doing research related to code I could borrow or base the rewrite on, I discovered that there was already an existing backup package with the name kbackup, so I decided to change the name to Cedar Backup instead.

Because I had become fed up with the prospect of maintaining a large volume of Perl code, I decided to abandon that language in favor of Python. ^[1] At the time, I chose Python mostly because I was interested in learning it, but in retrospect it turned out to be a very good decision. From my perspective, Python has almost all of the strengths of Perl, but few of its inherent weaknesses (I feel that primarily, Python code often ends up being much more readable than Perl code).

Around this same time, skyjammer.com and cedar-solutions.com were converted to run Debian GNU/Linux (potato) ^[2] and I entered the Debian new maintainer queue, so I also made it a goal to implement Debian packages along with a Python source distribution for the new release.

Version 1.0 of Cedar Backup was released in June of 2002. We immediately began using it to back up skyjammer.com and cedar-solutions.com, where it proved to be much more stable than the original code.

In the meantime, I continued to improve as a Python programmer and also started doing a significant amount of professional development in Java. It soon became obvious that the internal structure of Cedar Backup 1.0, while much better than kbackup, still left something to be desired. In November 2003, I began an attempt at cleaning up the codebase. I converted all of the internal documentation to use Epydoc, ^[3] and updated the code to use the newly-released Python logging package ^[4] after having a good experience with Java's log4j. However, I was still not satisfied with the code, which did not lend itself to the automated regression testing I had used when working with junit in my Java code.

So, rather than releasing the cleaned-up code, I instead began another ground-up rewrite in May 2004. With this rewrite, I applied everything I had learned from other Java and Python projects I had undertaken over the last few years. I structured the code to take advantage of Python's unique ability to blend procedural code with object-oriented code, and I made automated unit testing a primary requirement. The result was the 2.0 release, which is cleaner, more compact, better focused, and better documented than any release before it. Utility code is less application-specific, and is now usable as a general-purpose library. The 2.0 release also includes a complete regression test suite of over 3000 tests, which will help to ensure that quality is maintained as development continues into the future. ^[5]

The 3.0 release of Cedar Backup is a Python 3 conversion of the 2.0 release, with minimal additional functionality. The conversion from Python 2 to Python 3 started in mid-2015, about 5 years before the anticipated deprecation of Python 2 in 2020. Support for the 2.0 release was dropped on 11 Nov 2017. All users should migrate to the 3.0 release at this time.