1 |
Le samedi 01 mars 2014 à 10:06 -0600, William Hubbs a écrit : |
2 |
> On Sat, Mar 01, 2014 at 06:48:54AM +0000, Steven J. Long wrote: |
3 |
> > On Fri, Feb 28, 2014 at 09:31:08PM -0600, William Hubbs wrote: |
4 |
> > > On Fri, Feb 28, 2014 at 09:47:05PM -0500, Wyatt Epp wrote: |
5 |
> > > > But let's be real here: if I install something and |
6 |
> > > > want to configure its system-wide bits, the first place I go is ALWAYS |
7 |
> > > > /etc. When I don't find it there, with the rest of the system config |
8 |
> > > > files, my day gets a little worse and I lose a bit of time trying to |
9 |
> > > > interrogate a search engine for the answer. And that's annoying. |
10 |
> > > > That sucks. |
11 |
> > > |
12 |
> > > This hasn't changed. |
13 |
> > > The configuration files these packages are putting in /lib are not |
14 |
> > > meant to be edited; they are the package provided defaults. If you want |
15 |
> > > to override one of them, you do that in a file with the same path and |
16 |
> > > name in /etc, like I mentioned in another message in this thread. |
17 |
> > |
18 |
> > The problem, as has been explained many many times, is that the rest |
19 |
> > of the config is somewhere random on the system. But you knew that, |
20 |
> > right? You were just telling a half-truth, effectively. |
21 |
> |
22 |
> No sir, I was not telling a half-truth. |
23 |
> |
24 |
> If the default configuration is stored in /lib/udev/rules.d for example, |
25 |
> and you can override that default by dropping files of the same name in |
26 |
> /etc/udev/rules.d, I don't see what the concern is. |
27 |
> |
28 |
> > I for one prefer a distro to do a bit of work and make my life easier, |
29 |
> > since it makes life easier for everyone who uses the distro. Why the |
30 |
> > hell should I care if some bindist can't etc-update? WTF does that |
31 |
> > have to do with Gentoo? |
32 |
> |
33 |
> With this method, you don't need to etc-update, so I would say that in a |
34 |
> way this is easier. Your system-admin-provided files in /etc are not |
35 |
> owned by the packages, just the files in /lib are. |
36 |
> |
37 |
> > If I wanted a shitty distro that didn't bother to do anything at |
38 |
> > all, I'd use LFS. At least they don't pretend, then fall over themselves |
39 |
> > to do a crap load of work rather than admit a mistake; that hey, y'know |
40 |
> > what? Some of those things from 30 years ago were a damn good idea, |
41 |
> > and maybe just maybe, they worked some of these issues out back then, |
42 |
> > so we could stand on their shoulders instead of digging through |
43 |
> > their garbage. |
44 |
> |
45 |
> I'm not totally against keeping things from the past. It is just a case |
46 |
> of evaluating those things and seeing whether they are still relevant. |
47 |
|
48 |
I think the biggest issue here is that if the filename changes or the |
49 |
setting that is overridden changes, then end-user or sysadmin is the one |
50 |
that will suffer from settings not being applied and not knowing why. |
51 |
|
52 |
This already happened with systemd/udev and net rules for example and I |
53 |
am pretty sure in a couple of other packages but I have no other |
54 |
examples on the top of my head. |
55 |
|
56 |
Sure at some point you have to make things evolve but this upstream |
57 |
solution simply isn't nice for its users. |
58 |
|
59 |
-- |
60 |
Gilles Dartiguelongue <eva@g.o> |
61 |
Gentoo |