1 |
On Sat, Aug 3, 2013 at 7:30 PM, William Hubbs <williamh@g.o> wrote: |
2 |
> On Sun, Aug 04, 2013 at 01:49:46AM +0300, Alon Bar-Lev wrote: |
3 |
>> On Sun, Aug 4, 2013 at 1:38 AM, William Hubbs <williamh@g.o> wrote: |
4 |
>> > Actually the plan is to generalize it so that it works with other init |
5 |
>> > systems. Right now it is very tightly integrated with OpenRc, but there |
6 |
>> > is interest in changing that, so adding openrc to the name would be |
7 |
>> > misleading eventually. |
8 |
>> |
9 |
>> OK... so gentoo-networking? or just come up with own name? best-networking? |
10 |
> |
11 |
> I don't know about "best-networking". ;-) One reason we are splitting it |
12 |
> out also is so it can have its own development/release cycle; the |
13 |
> oldnet stuff is responsible for about 1/3 of the bugs against OpenRc |
14 |
> right now, and it is very difficult to test because of all of the |
15 |
> possibilities. |
16 |
|
17 |
You and I have had this talk more times than I can remember at this |
18 |
point. Using the name "oldnet" sucks and was one of the worst choices |
19 |
possible. Looking through our IRC chats, I had also suggested |
20 |
gentoo-networking. Its really a shame because this dependency based |
21 |
networking is really one of the real strengths of Gentoo and its |
22 |
really just being given the cold shoulder. Our dependency based init |
23 |
system was one of the reasons many people used Gentoo back in the day |
24 |
(besides the zomg its source its faster use -O6 crowd). |
25 |
|
26 |
In our discussions I've suggested trimming back the support in the |
27 |
networking scripts from all the various options (they support 4 DHCP |
28 |
clients for example) and slowly push these scripts to be installed by |
29 |
the package and maintained by the package maintainer (e.g. the pump |
30 |
script can be managed by the net-misc/pump maintainer and so on and so |
31 |
forth). Give them a name (systemd did by calling these snippets unit |
32 |
files) like OpenRC net snippets (Don't let me come up with names, I'm |
33 |
not good at that part, just ask my co-workers) and maintain an "API" |
34 |
for them. |
35 |
|
36 |
> |
37 |
>> However, I do not understand how you can port it without changing the |
38 |
>> notations... or lowering features... example: rc_net_*_provide, |
39 |
>> rc_net_*_need, or the rc_config, rc_use, rc_net_*_provide="!net" |
40 |
>> etc... |
41 |
>> |
42 |
>> Do you think systemd users can understand that /etc/conf.d/net is |
43 |
>> actually a shell script? I hope this is not going to be eliminated, as |
44 |
>> I use it a lot. |
45 |
> |
46 |
> Robin is going to do most of it I think; I'm not exactly sure what he |
47 |
> has in mind. |
48 |
> |
49 |
> William |
50 |
> |
51 |
|
52 |
If we lose that flexibility and configurability then just give up on |
53 |
OpenRC right now cause its dead because all interesting features are |
54 |
gone and it'll just become an inferior init system that needs to be |
55 |
replaced. |
56 |
|
57 |
-- |
58 |
Doug Goldstein |