[PLUG] Probing init scripts...

[Michael C. Robinson]

I've written a series of  bash script for selecting network boot
images.  I have a book on bash, but I'm running into a
lot of syntax I don't recognize and more than that I wish there 
were some notes on what these init scripts were written to do
in more detail than I currently have.  I can pick out some 
blocks of code and see some of the logical organization, but 
it's slow going.

Doing research online on how to create one's own nfs-root 
system I've taken a recommended first step of installing a 
normal Linux distribution.  Leaving all the extra package 
groups out in the Redhat 7.3 install I've gone further and 
removed even more packages I don't need so at least I'm 
working with a more minimized environment.  One thing I really 
don't like about Redhat is how spaghettied some packages including 
cyrus-sasl and openldap are with other packages.  One thing 
I don't get is why the files under etc for the large part 
aren't connected to the package their respective programs 
are installed from.  Another thing I don't get is why Redhat 
has a predefined password list instead of having a package 
add users and groups for itself when it is installed?  It 
would be nice to have some notes on what the default vendor 
accounts are for, etc.  That reminds me, I haven't trimmed 
any unneeded passwords and groups yet.  

I read somewhere that I should mount boot read only for better 
security, I'm not sure that's a good idea having caught a boot 
time error that System.map could not be recreated.

So far I've forced this package off:

     openldap

And I've removed:

     sendmail, procmail, mailcap, lokkit, lilo, ipchains, grub, gpm, 
at, and apmd.

I figured that it's ok to delete maillog with sendmail removed.

I'm down to 125 packages.

I want to remove cyrus, but I keep getting errors that libuser needs it.
I should probably remove the kerberos packages.  I wonder what else
I can take off intending this to be a system to run over nfs that
provides access to locally installed systems on a client's machine for
backup and restoration operations?  I should probably just link the
password files on the server instead of keeping the local password files
this system has now once I get it moved over.

I've tarred up X11 under etc and removed the directory from there which
shouldn't have any effect since X isn't installed.  I've done the same
with the ppp directory.  I've also deleted hosts.allow and hosts.deny.

Somewhere I read that config files for stuff that isn't installed can
be safely deleted and that they will be recreated if you install the
program that needs them.  Is there an easy way to catch all unused
config files?  The fewer places I have to look in, the sooner I'll 
be able to understand the system I'm working with well enough to 
modify it for it's particular purpose.

Thanks to the vi tip I've been able to fix the initial installation log
so that I can compare it to a fresh list of the installed rpms to verify
which packages I have removed.

Are anacron and vixie-cron different packages?  I should probably take
anacron off.  

Ultimately figuring out what's needed and what isn't a custom install
image wouldn't be a bad idea.  I guess I'd have to replace some standard
packages as it's looking like a number of files under /etc just aren't
needed.  Evidently the secret is to learn about anacron when it comes
to building a new installer for Redhat.

    --  Michael C. Robinson