rollback capability for linux config files and scripts

[Satish Balay]

On Mon, 10 Mar 2008, Rahul Nabar wrote:

> I am a newbiee to versioning systems and was looking at the best way to
> solve a problem I was facing:
> 
> Frequently, when I modify my scripts or config files on my Linux box I'm
> afraid I'll break something and hence make a manual backup copy. This only
> gets me a manual, rudimentary rollback capability. I was looking into using
> cvs / mercury / subversion to help me get this a little more efficient and
> safer.
> 
> The problem is that these files are peppered all accross my directory
> structure. There might also be a chance that I have two different files at
> different places in my dir structure but both with the same name. { Say,
> ~/gnuplot/script.sh ~/bin/script.sh)
> 
> What's the best versioning strategy for such a setup?  I browsed through the
> documentation of many of the versioning systems and the Mailing Lists. Most
> uses seem to be geared around "projects" each corresponding to a "repo".
> Will I have to twist my problem to fit that paradigm? Or is there a better
> way? I probably don't want to end up with 50 different projects; especially
> since this is entirely an artificial meta-structure for me.
> 
> Or am I trying to use the wrong tool here? I really don't need any
> collaboration requirement; just a historical rollback capacity. Probably
> need only a tiny fraction of the power of todays versioning tools!
> 
> Any suggestions?

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..

Alternative is to place all the relavent files into a single mercurial
[or any other tool] repository. And then create soft links to the
correct locations. This works fine as long as all files are in a
single user account.

Satish