David G. Miller wrote:
Since you edit your configuration files to aa great degree, do you
just replace the files from the new install or go through each to note
format changes? Upgrades leave rpmnew or rpmsave files, so short of
losing out on technological changes, what would make one be better
than the other. Merging config files from rpmsave or rpmnew files
should serve the same function.
I tend to work from both ends. System stuff in /etc gets the new
installation configuration. After the install is done I try to figure
out what needs to be changed and "bring forward" whatever I customized.
This means I tend to maintain a stable system while I try to figure out
what else needs to be changed.
Sounds like a wise practice. I was wondering what putting back the
previous configuration files would gain for the newly installed system.
comparing saved backups for upgraded configurations sounded pretty much
the same outcome. edited config files are usually saved on upgrades, so
comparison with these files would be very similar to saved backup
versions, except a new install will not cue you into the changed files
as easily as with an upgraded installation.
User stuff gets restored and then fixed if its broken. So, on the
laptop I mentioned, I edit /etc/sysconfig/iptables, /etc/fstab, etc. to
bring them back to where I want while the various user rc and
configuration files just get dropped into place. If something breaks, I
back out the change and go from fresh. I still have the original config
file as reference.
Less system critical files should be alright to do this with. I guess
with no complaints, things go alright.
BTW, speaking of CDs... it appears that a default install only requires
the first three CDs. I didn't even bother to burn CDs 4 through 6 for
the x86_64 install. I had the ISOs downloaded in case I needed them but
didn't. For everything beyond a default install, I just pulled stuff
with yum.
I believe there was a lot of work by the distribution developers to get
this to work for default installs. It sure beats needing disk 5 for one
random package which gets placed there, if you only want a default
system installed and build it up from there.
The other problems are obsoletion and unsupported packages.
Rhetorical question: what should an upgrade do if a user program is
now obsolete and the replacement is one of several different
programs? Unsupported packages are even worse for a distro like
Fedora or RHEL. I run xmms-mp3. What should Fedora or Red Hat do
when I upgrade? Hint: their lawyer may disagree with your solution.
Leave it broken, the application of updates after the install is
finished should allow the program to function again as intended.
Works well for user apps but I lived through the evolution of ipfwadm ->
ipchains -> iptables. Need to be careful with system stuff. It would
be nice to see core functionality supported for upgrades even if every
oddball app isn't. One of the arguments against supporting upgrades is,
"if it ain't broke, don't fix it." That is, once a release supports a
platform, why change. As with my laptop example, there are good reasons
to upgrade from an OS release that only marginally supports a hardware
platform to one that fully supports it. Let's hope somebody at
Fedora/RH listens.
I remember ipchains but not ipfwadm. Anyway system critical packages
should be upgraded or left intact when upgrading a system.
Jim
Cheers,
Dave
--
"I remember when I was a kid I used to come home from Sunday School and
my mother would get drunk and try to make pancakes."
-- George Carlin
