-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Heh, I just forgot the IMHO most important idea that came up. :-)
What about different ifile branches, a generic one and one for each distro
we are able to support. The generic branch would use boot-time
configuration in /etc/initng/bootconf.d, configuration settings could be
retrieved with a fast C program that is called from scripts. These generic
calls would be distinctive and not very special, beeing easlily
replaceable.
I'm currently working on that C program (I call it "bootopt"), because I'm
going to use it with my own ifiles branch. I'm going to stick with the
Bernstein Way, which basically means that every option is a file. If a file
existis, the option is set. If the option takes an argument, we read _one
line_ from the file. Nobody can make any mistakes that way, syntax errors
are gone forever.
Bootopt takes the follwing switches:
- -b (boolean) Exit with 0 if the option is set (the file exists)
or 1 if not. Does not output anything.
- -g (get) Prints out the option set (or nothing if the file is
empty), returns 0 if the option is set, 1 otherwise. This is the
standard behaviour.
- -s (set) Sets the option to a specified value or to \"true\" if
used in conjunction with -b.
In the distro-specific ifiles, the bootopt call could either be directly
replaced with the corresponding call (mostly including some shell script)
that is used on this distro, or bootopt could evolve to a transparent layer
for alot of configuration schemes.
Again, tell me what you think. :-)
On Saturday 27 October 2007, Eric MSP Veith <[EMAIL PROTECTED]>
wrote:
> Hello!
>
> While customizing my ifiles installation, some ideas popped into my mind
> I want to share with you and about which I'd like to read your comments.
>
> I'd like to introduce a virtual service that marks the earliest point at
> which the bare system is usable (i.e., a root login is possible). The
> would be similar to the rc.S script in a sysv environment. Other services
> and daemons could depend on this virtual service. This would (a) clarify
> the "the system is now more or less up" point we now implicitly mark with
> system/bootmisc (which is, in my opinion, not a good style), and it (b)
> could speed up the boot process a little more, because services that are
> not that important for a system initialization could be started more in
> parallel with daemons.
> Important for the bare system are udev, lvm/raid configuration, (local)
> networking (see below), mountfs, clock (because sometimes, when the clock
> service is run to late, I encountered that DHCP stopped working because
> the PID file had a modification time in the future) and getty.
> I'd suggest "virtual/basesys", "virtual/sysinit" or something similar for
> this.
>
> At the moment, we have a virtual/net service. However, when setting up
> NFS environments (for /home and /usr), I found that virtual/net is called
> to late. I'd like to split it up into virtual/lan und virtual/wan
> (virtual/net may depend on both). "virtual/lan" would include static
> ethernet configurations and DHCP, while virtual/wan would be for PPP,
> PPPoE and others.
> The goal is to divide networking into a part that must be present early
> in the boot process for NFS and the like, and other networking services
> that are not so important for the bare system, like an internet
> connection made via the pppd. But a firewall script could continue to
> rely on virtual/net, thus having both eth0 and ppp0 present for the
> iptables setup.
>
> These ideas are rather raw; I'm looking forward to your comments and
> suggestions.
>
> Eric
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)
iD8DBQFHIntzfkUtd7QNU/sRAnRMAKCre+CXodjPO/5Z1BIb7aqwF8O+pgCdHVP5
k4rDyEnxOa8C7n75DZYrEsY=
=+Y1e
-----END PGP SIGNATURE-----
--
_______________________________________________
Initng mailing list
[email protected]
http://jw.dyndns.org/mailman/listinfo/initng
