1 |
>>>>> On Thu, 1 Mar 2012, Fabian Groffen wrote: |
2 |
|
3 |
> I think that discussion can be split into |
4 |
> 1) if a UTF-8 profile should be set as default, and |
5 |
> 2) what UTF-8 profile that would be |
6 |
|
7 |
> Just to give some hints/ideas here: |
8 |
|
9 |
> For 1), for many people it's a waste, and it's not POSIX (== C |
10 |
> IIRC), so should we really do that. Also, how about this changing |
11 |
> default for existing systems. Why not update the docs to strongly |
12 |
> suggest to set an UTF-8 profile instead. |
13 |
|
14 |
We shouldn't set LANG to anything else than POSIX (or C), otherwise |
15 |
there will be undesired side effects affecting for example number |
16 |
formatting, collating sequence, units of measurement, and paper size. |
17 |
|
18 |
If we change the default, we should use a less intrusive approach and |
19 |
change LC_CTYPE only, which is the relevant variable that controls the |
20 |
character encoding functions. |
21 |
|
22 |
> Then, 2) is almost moot. If you would default to an UTF-8 profile, |
23 |
> I think the only sensical option is en_US.UTF-8 (not en_GB). |
24 |
> However, if you only strongly suggest in the installation docs, you |
25 |
> can get away from this problem, since the user has to choose his/her |
26 |
> preference in that case anyway. Just a note that leaving |
27 |
> en_US.UTF-8 always in is a good idea would suffice, I guess. |
28 |
|
29 |
I agree that en_*.UTF-8 is a reasonable setting. I think for LC_CTYPE |
30 |
it doesn't even matter if it's en_US or en_GB. |
31 |
|
32 |
Ulrich |