rollback capability for linux config files and scripts

[Matt Mackall]

On Tue, 2008-03-11 at 18:29 -0500, Rahul Nabar wrote:
> >If its just keeping track of changes to a few files spread out [in
> >different directories], I would just use RCS. Its primary purpose was
> >to track individual files changes anyway..
> 
> Thanks Satish for your helpful comments.
> Actually Satish's  pointing me to RCS led me to another article on a
> similar topic: http://www.linux.com/articles/52568
> 
> It seemed to me "Bazaar OS" (http://bazaar-vcs.org/) was the best tool
> for my task ( cvs, subversion, mercurial and rcs were the tools that I
> had a quick peek at) What tilted the balance in favor of "Bazaar OS"
> was that it seemed the only tool that didn't pepper my dirs with
> the .cvs .hg etc. files. It seems to keep all its bookkeeping hidden
> away from view in its own folder. I liked that especially since I
> would be versioning a large number of seperate files and dirs. Just my
> preference, I guess.

Now you can expect to get dozens of people telling you that Mercurial
has one .hg directory per repository.

> Only a minor annoyance but still.  Unfortunately, I did not realize
> that mercurial was capable of similar stuff
> ( http://www.selenic.com/mercurial/wiki/index.cgi/TrackingEtcEtc
> Thanks Mads! ). I'm guessing the mercurial implementation is probably
> more efficient that a python implementation so perhaps I will switch
> later if performance becomes an issue.

Mercurial is also mostly in Python, but the most performance-sensitive
parts (and there are only a few) are in C.

>  BTW, does doing a hg init on / add any performance drags? In the
> sense, that this is a HUGE filesystem for most people. Just curious.
> (most documentation seems to be about "hg init" on a specific project
> repo.)

cd /
hg init
echo ".*" > .hgignore # only track files that are explicitly added
hg add /etc/apache/*

-- 
Mathematics is the supreme nostalgia of our time.