Fwd: Re: [LBo] Permanent /home partition

[Randy Kramer]

Thanks for taking the time to reply so carefully.

My first general response is:

How would you like it to be, say 100 years from now?

How hard is it to start making the change?  

Pretty simple: 

   * You get a few people (FHS and LSB) to make the change a goal.

   * As new applications are written, or old ones are maintained, they are 
written to support the new approach (look first in /data/<user> 
(or /<config>/user)--if the necessary files are not found there (or if the 
directories don't exist) fall back to looking in /home/<user>.

Maybe it could be a goal for 2032?

Randy Kramer

----------  Forwarded Message  ----------

Subject: Re: [LBo] Permanent /home partition
Date: Wednesday 07 June 2006 10:10 pm
From: Brian Reichert <rignes at ptd.net>
To: qna at linuxbasics.org

What follows is my proverbial "2 cents" on the matter:

I just poked about the FHS [1] regarding home directories [2] and see that it
specifies things exactly as they currently are (no surprise there).  I've
thought about what you suggest, to separate config and user data, and I'm not
sure I agree that the advantages are significant enough in my case to be
 worth the trouble.  Actually, I rather like having one place with it all in
 because a single tar command makes a nice backup of all my stuff.  Sure,
 restoring just the file you want can be a bit of a pain that only happens
 rarely (hopefully!) and I consider it bad practice to restore directly back
 into your home directory because you may accidentally overwrite something. 
 I always restore to a temporary location then copy over the files I wish to
 restore just to be safe.

To backup your data all by itself with out any configuration files can simply
be done by putting your data in a "data" directory within your home
 directory. For instance, in my case, I have /home/rignes/Documents in which
 I store all my data in.  Most everything else outside of that is extra stuff
 that doesn't really /need/ to be backed up.  I also, by the way, have a
 no_backup directory in there to keep things that I don't want to backup. 
 Things like movie files, or my install of Unreal Tournament.  My point
 being, there are other ways that are simpler to backup just data without
 changing the standard.  Also, to backup just your "." files you could just

$ tar cvjf dotfiles.tar.bz2 $(ls -A1 | grep '^\..*')

which should just get dot files and dot directories and their contents.  BTW,
this even avoids trying to backup the . and .. directories.  Oh, if anyone is
interested I can break down what the parts of this command mean and do.  Just
let me know.

Now, as far as running multiple distributions on a single machine and sharing
the home directory I can't really comment on that.  I always use a spare old
machine here to try out a new distro.  Though, I'm willing to bet that the
number of people overall who have a need to run multiple distributions on a
single machine compared to those who don't is small.  I think that a high
number of them exist on this list though.

Now that my personal opinions are out of the way let me toss out a thing or
 two to think about before anyone starts contacting the folks who manage the
 FHS. Obviously, changing something in the file system standard can be a
 painful prospect until it is complete and widely accepted.  Lets think about
 the big, overall picture and really try to answer a few important questions
 to keep things in perspective.  Such as:

How many organizations/users would be interested in such a change?  Is the
number small?

What percentage of people/organizations really /want/ to have two separate
locations for their data?  Many probably enjoy the single location concept
because it is simple.

How many software packages would need to be rewritten or recompiled and
packaged?  What of those that will /never/ be rewritten or repackaged but are
still essential to the users/organizations environment?  What provisions must
be in place to deal with these?

How many people, over all, really want to run multiple distros on the same
system while sharing their home directory between each distro?

It's important to list off all the pros and cons to such a change and to
 think about things in great depth when changing something this fundamental. 
 Once all this analysis would be done the final question would be, "Do the
 benefits of this change effect enough people and/or organizations to be
 worth the difficulties of implementation?"  Let me also mention the KISS
 theory (Keep It Simple Stupid) and say that often the simplest solution is
 the best one. Avoiding unnecessary complexity is a good thing.  An overly
 complex system is easier to fix, maintain, troubleshoot, and tends to be
 more stable than a complex system (think less points for failure).

I have a feeling that in this case the needs met by splitting ones home
directory applies only to a small niche and that the difficulties of
implementation are far more then the benefits gained.

Though, I may suggest that perhaps this concept could
become a project and people who have this need could work together to create
either some software, or highly detailed howto's to split out home
 directories.

  1. http://www.pathname.com/fhs/
  2. http://www.pathname.com/fhs/pub/fhs-2.3.html#HOMEUSERHOMEDIRECTORIES

--
Brian Reichert
http://rignesnet.tzo.com

The past always looks better than it was.  It's only pleasant because
it isn't here.
		-- Finley Peter Dunne (Mr. Dooley)

-------------------------------------------------------