Colin Walters wrote:
On Fri, Sep 12, 2008 at 5:59 PM, Matthew Woehlke
<mw_triad@xxxxxxxxxxxxxxxxxxxxx> wrote:
Colin Walters wrote:
We're going to be removing the legacy non-X system consoles by default
in the long run.
Um... what happens then when X is broken?
What happens when the linux kernel is broken?
...then you boot from a disk with a bootable kernel.
What happens when /bin/sh is broken?
Hmm... then you probably need a rescue disk to replace /bin/sh.
What happens when NetworkManager is broken?
You fix it :-). You've degenerated from errors that result in an
unusable system (meaning: cannot boot to a command prompt from which, at
the least, basic repair tools are available) to errors that are
potentially mere annoyances.
Let's go back to my question:
What happens when X is broken?
Then you boot in non-X mode.
Oh, wait. You want to remove non-X mode. You want me to have to go find
a rescue disk, just because 'yum update nvidia' wasn't such a good idea?
No, thanks; I'd rather have X fail to start and dump me at a normal
console from which I can fix the problem *without rebooting*, much less
needing to dig up a rescue disk :P.
Currently non-functioning X is like non-functioning network; annoying,
but not crippling (kernel still works, shell still works, still possible
to run commands, do debugging, etc., without rebooting or a rescue
disk). You're proposing to make it a system-crippling problem. I will
suggest that making a known-fragile component *required* for a
functional system, including any chance of repair (short of a rescue
disk), is the worst idea I've heard in a while. (But I /have/ heard it
before. It's called "Windows". YTH do we want to copy /that/?)
...not to mention that X is bloated and completely unneeded overhead on
servers. Even Microsoft realizes this; I hear 2008 has a non-GUI mode
for exactly this reason.
Also, one thing I would like to see Fedora install by default is a
compressed recovery image, rather than just multiple kernels.
That would help, but would still require rebooting all the time until X
is working (anyone say "PITA"?). Currently, "bad kernel" is the only
situation that needs a reboot to test if it works. (Well, non-working
/bin/sh depends on why it is non-working; if it's a bad binary, you can
try to run it as a subshell, so no reboot needed. If it's an interaction
with other bits of the system, it might fall into the 'needs reboot to
test' category. But non-working X currently is comfortably NOT in the
'needs reboot' category.)
--
Matthew
What? This signature /again/?
--
fedora-devel-list mailing list
fedora-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-devel-list
