1 |
On 15-05-2008 10:12:01 +0200, Markus Duft wrote: |
2 |
> Hi |
3 |
> |
4 |
> I just fixed the openssh server for interix (yeehaa), and for the |
5 |
> purpose of starting this whole thing without worrying about keys and |
6 |
> such, I would have really liked to use the init.d script to fire it |
7 |
> up. The init script still would require some eprefixification, but the |
8 |
> bigger problem is, that runscript is missing in baselayout-prefix... |
9 |
> Are there any thoughts in that direction? |
10 |
|
11 |
Yeah, search bugzilla and the -dev list on gsoc prefix ideas. |
12 |
|
13 |
> I can think of three scenarios on how services in prefix _could_ work: |
14 |
> |
15 |
> 1) all init.d scripts are run when entering the prefix, thus called by |
16 |
> some bashrc script. |
17 |
> 2) one single command is registered as a service in the host system |
18 |
> (called the "prefix-service" or so *lol*), which kicks on all |
19 |
> other init.d's from the prefix. |
20 |
> 3) each service has to be started manually by running the init script. |
21 |
|
22 |
All three have been discussed on -dev I think. |
23 |
|
24 |
> Any thoughts about this? What about openrc? I must admit, that I right |
25 |
> now have just a very shallow knowledge of all this... still I feel |
26 |
> that sooner or later we'll need _some_ facility to run init.d scripts |
27 |
> - even if it's manual only. |
28 |
|
29 |
openrc is our only hope/option. The bastard baselayout we have right |
30 |
now is REALLY old (it dates back from the early dark ages of 3rd |
31 |
generation Prefix, when vikings were still hyjacking packages and |
32 |
machines...) |
33 |
|
34 |
> A little background on why I'd need this (I could of course do what I |
35 |
> want without init scripts, but it would be more painfull): we have an |
36 |
> automated setup for new interix machines in our company. Since we use |
37 |
> prefix as the "real" system, and interix is just there to give us the |
38 |
> very basics while bootstrapping (and the libc), it would be really |
39 |
> cool to run sshd from prefix, instead of installing some binary |
40 |
> version with closed-source patches applied (urgh...) |
41 |
|
42 |
You don't have to expect any straw laid on your road from me (that |
43 |
probably isn't English, but maybe you get the point ;) ), it is just |
44 |
that I don't have the time/interest/need for it. I do see a need for |
45 |
it. However, if you do start working on this, please be open about the |
46 |
process, discuss, and take some of the bugs on it into account. I |
47 |
personally feel that the last thing we want with a thing like that is to |
48 |
have a hacky "just works" solution. Roy (UberLord) Marples already |
49 |
added some stuff that we need in OpenRC. |
50 |
|
51 |
I would prefer a single point in the system where you can "switch |
52 |
on/off" the prefix init system. That is, like for OpenRC when the |
53 |
system boots or shuts down. This need not to be related to the physical |
54 |
boot/shutdown. I think it needs to be a one-liner (a script, or a |
55 |
command built into openrc, maybe just "prefix-init 3" or something |
56 |
similar) that each user can hook in such a way that he/she prefers. For |
57 |
me it would be sufficient to start the stuff after I booted my machine |
58 |
(every once in a long while usually), a sysadmin might want to make it a |
59 |
/etc/init.d/prefix-init script... Whatever. I think the number of |
60 |
tastes are unlimited here. So let us just deliver the "magic button" to |
61 |
start and stop the entire system (for the runlevel stuff, rc-update add |
62 |
default bla stuff), and of course for many users |
63 |
`${EPREFIX}/etc/init.d/postgresql start` will be enough, so they don't |
64 |
need any automatic starting at all :) |
65 |
|
66 |
|
67 |
-- |
68 |
Fabian Groffen |
69 |
Gentoo on a different level |
70 |
-- |
71 |
gentoo-alt@l.g.o mailing list |