I have been keeping my LaTeX files under version control since I was introduced to AUCTeX and I realized that a good way to approach the "evolutionary" approach to software (this includes text) is to use a revision control system.

I started using that with Emacs 21 (the first version of Emacs that I used was Emacs 19.something which I can't remember) and the revision system that was most easily available was RCS. Its integration with Emacs was very good and also easy to use: type C-x v-v and a check in is performed (or a check out, depending on the state of the file). You would be asked for a changelog for that file and that was it.

Now, on the other hand, there is a myriad of options and one that looks very nice to me (apart from not dealing with metadata, but that's OK with me for the moment) is Linus Torvald's git. The very nice thing with git (or many other such systems) is that the interface with Emacs is exactly the same: C-x v-v. That's the only thing you need to remember to work well with, say, your text project. (I don't actually know very much of the management workflow under Emacs, but I also use the command line a good amount and it serves me well so far).

But how can I convert my RCS "repositories" to a proper git repository? Well, git comes with a subcommand called cvsimport, meant to be used with CVS and this does help, since a CVS repository is essentially just a collection of RCS files. As many people have migrated along the path from RCS to CVS to Subversion to git, and I did not intend to convert the repository in multiple steps, I am documenting what I did here.

So, if you have the ,v files that RCS generates (the "repository" files for each file tracked by RCS) spread in a tree and you would like to convert that tree to a git repository for replication, backup and "time machine-like" support, you can mimic a basic CVS repository that can be converted easily with the following simple steps:

• Create a new directory hierarchy somewhere, denoted here by cvs-source-dir. This will play the role of a CVS repository which git cvsimport will convert.
• Under cvs-source-dir, create the directory tree as you would like it to be visible by git.
• As an immediate subdirectory of cvs-source-dir, create a directory named CVSROOT, that is cvs-source-dir/CVSROOT.
• Each immediate subdirectory of cvs-source-dir will play the role of a CVS module. Just think of that as a project under control of the version control system.
• Think about the ,v files as your regular files and put them under the cvs-source-dir tree, as you would like them to be.
• Create the directory where your git repository will reside, denoted here by git-repo.
• Run git cvsimport -v -a -A -k -m -d cvs-source-dir -C git-repo cvs-module

That's it. It seems to perform a very good job of changing a very old system to something that is modern and with many features (and is distributed).

You may also want to read about the way to convert the ids of commits that RCS/CVS register to proper (people) names, which git uses. Read the git-cvsimport manpage for more details on this and other options mentioned above.