1 |
On Tuesday 27 September 2005 02:59 pm, Donnie Berkholz wrote: |
2 |
> Ciaran McCreesh wrote: |
3 |
> | Ok, for those who don't remember... |
4 |
> | |
5 |
> | In the good old days, syslog-ng was the default. Then, for one of the |
6 |
> | releases, the x86 team accidentally included metalog instead of |
7 |
> | syslog-ng on the distfiles CD, so the default was changed over to make |
8 |
> | the x86 networkless install documentation work (at the expense of the |
9 |
> | archs that got it right). So, it's not a case of moving away from |
10 |
> | metalog at all, it's a case of undoing a temporary change. |
11 |
> |
12 |
> In the bad old days, metalog was the default. Then it got switched to |
13 |
> syslog-ng because people got confused as hell by metalog's buffering, |
14 |
> and there was no reason not to use a more capable logger that's |
15 |
> compatible with the traditional sysklogd, minus the annoying feature of |
16 |
> metalog. |
17 |
|
18 |
ok, i thought that's how *i* remembered it :) |
19 |
|
20 |
> FWIW, I think metalog's buffering disappeared a while back, at least by |
21 |
> default. |
22 |
|
23 |
yes, default metalog config file is unbuffered (and has been for a while), and |
24 |
the init script lets you switch between modes now on the fly so you dont have |
25 |
to send the kill signals yourself. whether it's 'annoying' is certainly a |
26 |
matter of opinion :P, but from a technical aspect, buffering makes a lot of |
27 |
sense if the performance matters. |
28 |
-mike |
29 |
-- |
30 |
gentoo-dev@g.o mailing list |