1 |
On Thu, 05 Jan 2012 16:50:45 +0100 |
2 |
pk <peterk2@××××××××.se> wrote: |
3 |
|
4 |
> On 2012-01-05 13:08, Alan McKinnon wrote: |
5 |
|
6 |
[snip] |
7 |
|
8 |
> > mdev has a much narrower scope where things are considerably more |
9 |
> > static. |
10 |
> |
11 |
> Currently it does have a more narrow scope, yes, but that can change, |
12 |
> no? Although I'm not entirely convinced that a userspace dev manager |
13 |
> is needed (yes, devfs on Linux was an utter failure but Solaris, Mac |
14 |
> OS X, *BSDs use it[1] and done properly in Linux it should work just |
15 |
> as fine)... |
16 |
> |
17 |
> 1: https://en.wikipedia.org/wiki/Devfs#devfs |
18 |
|
19 |
As I understand it, devfs had unfixable race-condition problems. |
20 |
Normally these things can be fixed with an architectural re-design but |
21 |
Greg then showed up with udev. It's in-kernel design is quite elegant |
22 |
actually. But udev then and udev now are likely very different beasts. |
23 |
|
24 |
And that's about my limit of things I know for certain with udev |
25 |
|
26 |
> > As for re-arranging the fs layout, I think it was Canek in the last |
27 |
> > thread that gave an excellent example of why this is needed. When |
28 |
> > devices hotplug, or need to become active early on in the boot |
29 |
> > process, they need to run code that can be located almost anywhere. |
30 |
> > It wouldn't be fun trying to get a wireless keyboard going when |
31 |
> > it's start-up script needs to get into /usr/lib/firmware and /usr |
32 |
> > isn't mounted yet. |
33 |
> |
34 |
> Yes, I understand the need for this but... how does a wireless |
35 |
> keyboard work under bios/firmware (*efi)? Never tried one and never |
36 |
> will... A computer without ports should handle such connections in |
37 |
> firmware (analogy: You don't need software to drive a cable). |
38 |
|
39 |
To you and I it might seem absurd, but I think out there in the |
40 |
marketplace it's quite a reasonable thing for a user to want. And if |
41 |
that example never happens, there's hundreds more than can. Point |
42 |
being, the code must be able to deal with such things. |
43 |
|
44 |
> > I do agree with collapsing the executable code in /usr into /, or |
45 |
> > having /usr on the root partition. A separate /usr/{,s}bin is pretty |
46 |
> > pointless and was never done for safety or maintenance reasons. It |
47 |
> > was done way way way back when disks were small and a convenient |
48 |
> > hack was to keep the OS on the boot device and user apps somewhere |
49 |
> > else on bigger but slower storage (which often was remote). |
50 |
> |
51 |
> Hm... I find it quite elegant and flexible with the separation of / |
52 |
> and it's various underlying directories. I guess we can agree on |
53 |
> disagreeing here... although, I'm a bit surprised to see you as an |
54 |
> admin defending the "new" way... Windows does have such a philosophy |
55 |
> with putting everything system related into a directory (\WINDOWS)... |
56 |
> Ultimately one can argue why use anything else besides Windows, it |
57 |
> does the job reasonably well. |
58 |
|
59 |
Oh, I'm not advocating doing it Windows style, I'm simply saying what's |
60 |
the point of the system itself having 4 locations for binaries when I |
61 |
never use that separation? I gain nothing from having a /bin and |
62 |
a /usr/bin. |
63 |
|
64 |
/opt and /usr/local/bin *are* useful so I'd keep those. Same with /var |
65 |
and all the other traditional separate mount points. |
66 |
|
67 |
|
68 |
> > If /usr is local, what really is the point of having it separate |
69 |
> > from /? Have you ever found a Linux system in any condition that |
70 |
> > could not start just because the stuff in /usr was available? I |
71 |
> > haven't. |
72 |
> > |
73 |
> > Even the split between bin and sbin is arbitrary. It's only there so |
74 |
> > that users can take sbin out of PATH and not have the screen |
75 |
> > cluttered with endless junk when they tab-tab. It makes much more |
76 |
> > sense to me to just have one single bin and lib location and shove |
77 |
> > everything into it. |
78 |
> |
79 |
> I'm not an admin of a large organization so what do I know... but, I |
80 |
> still can appreciate the flexibility and "tidyness" it[2] gives you |
81 |
> in a multi-user system. I also can see this from a security point of |
82 |
> view ("keep the cool toys from the children")... I personally like it |
83 |
> for my very local computer as well for the above reasons (flex./tidy). |
84 |
> |
85 |
> 2: https://en.wikipedia.org/wiki/Filesystem_Hierarchy_Standard |
86 |
> |
87 |
> What you are basically saying is that everything "we" have learned |
88 |
> about computer systems should be abolished and we adapt the |
89 |
> monolithic, "black box" philosophy of newish systems like Windows. |
90 |
> That's how I interpret what you're saying (yes, I do know hardware |
91 |
> has changed since the 60'ies but not that radically, IMO)... I tend |
92 |
> to think of Unix as "Lego" where you have lots of little bits with |
93 |
> clean(ish) interfaces with which you can build whatever you want.dual |
94 |
> |
95 |
|
96 |
Good analogy. I also like building systems from individual Lego bricks. |
97 |
I don't like having to build the bricks themselves first :-) |
98 |
|
99 |
Windows goes too far to the other extreme IMO. That OS seems to have |
100 |
largely abandoned control and there's not much in the way of |
101 |
structure. Too little control is just as bad as too much |
102 |
|
103 |
|
104 |
[snip] |
105 |
|
106 |
|
107 |
|
108 |
-- |
109 |
Alan McKinnnon |
110 |
alan.mckinnon@×××××.com |