1 |
On Mon, Jan 5, 2009 at 1:15 PM, Luca Barbato <lu_zero@g.o> wrote: |
2 |
> Fabio Rossi wrote: |
3 |
>> |
4 |
>> I'm proposing to reorganize the files related to Gentoo inside /var/lib. |
5 |
>> Currently we have this situation (at least on my system): |
6 |
>> |
7 |
>> /var/lib/eselect |
8 |
>> /var/lib/gentoo/enews |
9 |
>> /var/lib/herdstat/ |
10 |
>> /var/lib/module-rebuild |
11 |
>> /var/lib/portage |
12 |
>> |
13 |
>> The main dir should be something like /var/lib/gentoo, so I'd see all |
14 |
>> gentoo-related files as |
15 |
>> |
16 |
>> /var/lib/gentoo/eselect |
17 |
>> /var/lib/gentoo/enews |
18 |
>> /var/lib/gentoo/herdstat/ |
19 |
>> /var/lib/gentoo/module-rebuild |
20 |
>> /var/lib/gentoo/portage |
21 |
>> |
22 |
>> What do you think about? |
23 |
> |
24 |
> Looks quite wrong. I'd do /var/lib/gentoo/enews -> /var/lib/enews btw |
25 |
|
26 |
A few things, mostly relating to trying to convince me to fix this |
27 |
(being open source, you could easily fix it yourself). |
28 |
|
29 |
1) There is no right or wrong place to put data, specifically if we |
30 |
state that we do not follow standards (like the FHS) that may in fact |
31 |
make specific recommendations, and we have done so in the past. |
32 |
|
33 |
2) There is not a real problem (technically) with the current |
34 |
placement, aside from some people think it is odd. I am unaware of |
35 |
users that are trying to put /var/lib/ on an NFS mount but can't |
36 |
because of our scheme; or other ideas for which the current scheme |
37 |
limits users. |
38 |
|
39 |
3) Moving is costly. |
40 |
- We have to author code to migrate existing news repositories |
41 |
- We have to author code to check both places (at a minimum, to |
42 |
migrate the old repository, possibly at runtime, or to warn them of |
43 |
the old repository) |
44 |
- How long will this code remain in the codebase (forever?) |
45 |
|
46 |
4) The benefits are slim. |
47 |
- /var/lib is more orderly for a subset of users. |
48 |
- What else? |
49 |
|
50 |
The point is we make decisions all the time and some of them are bad |
51 |
and are expensive to fix and we have to live with them. |
52 |
|
53 |
The location was actually specified in the GLEP[1] so part of me asks |
54 |
why this was not brought up then (it was a design defect; and not an |
55 |
implementation defect). Data from software engineers tends to show |
56 |
that defects in the design phase that are caught in the maintenance |
57 |
phase tend to be very expensive to fix and I think there are bigger |
58 |
fish to fry in the package manager world; but feel free to send a |
59 |
patch as always (this is OSS after all). |
60 |
|
61 |
[1] http://www.gentoo.org/proj/en/glep/glep-0042.html#client-side |
62 |
|
63 |
> |
64 |
> lu |
65 |
> |
66 |
> -- |
67 |
> |
68 |
> Luca Barbato |
69 |
> Gentoo Council Member |
70 |
> Gentoo/linux Gentoo/PPC |
71 |
> http://dev.gentoo.org/~lu_zero |
72 |
> |
73 |
> |
74 |
> |