[totally OT] merits of backing up your server config in RCS (was: Re: [rfc] Windows symlink support)

[Jamie Wilkinson]

This one time, at band camp, Jari Aalto wrote:
>Jamie Wilkinson <jaq at spacepants.org> writes:
>
>> Both these use cases are more easily solved using management agents like
>> puppet or cfengine; pushing symlinks out from your RCS is a terrible way to
>> handle this.  Usually one requires a tool run after the RCS push to modify
>> the final layout.  In such a case I've found it best to let this tool do the
>> linking itself, and leave the symlinks out of revision control.
>>
>>>Some people version $HOME, which I have started playing around with.
>>>Mostly because I have 10 different machines I work on, at various sites
>>>(home, work, laptop, school, etc), where doing an NFS mounted home
>>>directory is not possible (kind of hard to do on an airplane), but it
>>>also sucks to try and keep my ~/.vimrc .vim/* in sync.
>
>CNERALIZED MANAGEMENT

This is totally not the forum for this, but...

>    People, CVS has been there for ages to keep separate hosts and 10
>    different user identities (root, cvsadm, wwwadm, $USER) well on synch
>    :-) It has been ideal for star-like system, where all hosts share the
>    same configuration (even if they are on different OS, you can still
>    separate "common" from site-specific).

Yes, I know how you can do config management for systems administration
using CVS, using Subversion, using bazaar, using whatever.  The point I was
trying to make and everyone seems to be missing, their vision blinded by
red-rage at the thought of no symlink support, is that using *only* a RCS
for server admin in the manner you describe is suboptimal for real sysadmins
in the real world.

Meanwhile, back in bzr-dev land, I repeat: I am not opposed to symlink
support in bzr if it doesn't break on windows.