Date: Sun, 20 Jul 2008 18:03:29 +0300 From: Danny Braniss <danny@cs.huji.ac.il> To: Peter Ross <Peter.Ross@alumni.tu-berlin.de> Cc: Doug Barton <dougb@FreeBSD.org>, Robert Noland <rnoland@2hip.net>, freebsd-current@FreeBSD.org, ticso@cicely.de Subject: Re: rc improvements (wanted?) Message-ID: <E1KKaRd-0008h9-Hl@cs1.cs.huji.ac.il> In-Reply-To: <20080720200229.T894@oldie.bigpond.com> References: <200807172056.08835.naylor.b.david@gmail.com> <487FCA89.2010308@FreeBSD.org> <20080718083725.97823be0tg13fn6s@webmail.leidinger.net> <20080718071806.GV62764@server.vk2pj.dyndns.org> <20080718122928.GD35340@cicely7.cicely.de> <4881795A.4070604@FreeBSD.org> <1216473536.1991.5.camel@wombat.2hip.net> <20080720003343.S23554@oldie.bigpond.com> <E1KKUpI-0004Ld-CR@cs1.cs.huji.ac.il> <20080720200229.T894@oldie.bigpond.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> On Sun, 20 Jul 2008, Danny Braniss wrote:
>
> > I went ahead with my idea - to reduce the list rcorder delivers, by
> > eliminating those that don't have ${name}_enable, and I opened a pandora
> > box :-)
> > - dummy dependency like SERVERS/LOGIN don't have ${name}_enable
> > nor should have.
>
> Two options:
> - Just eliminate the scripts that have the variable, and set it to "no"
> - A dummy entry in /etc/defaults/rc.conf with "# Don't overwrite this"
> (So only one who really knows what he is doing will try it - hopefully)
>
as usual I forgot one word,
don't have ${name}_enable
should have been:
don't have ${name}_enable="YES" (or true or 1)
personaly, I prefer self-contained, like
iscsi_enable=${iscsi_enable:-"NO"}
instead of adding it to /etc/defaults/rc.conf, though I can see the benefit.
what I'm trying to say, is keep the complexity to a minimum, less
foot-shooting :-)
> > - REQUIERE: xxx complains if xxx is not 'loaded' like in the case
> > of NETWORKING requirement of ppp which I don't have enabled.
>
> Solved by point 1?
>
> > - some scripts rely on the existance of a ${file} which is better than
> > the original /etc/rc which used to run mountd if /etc/exports existed,
> > but does not 'conform' to the ${name}_enable paradigm.
>
> The same?
>
> > - some scripts like abi don't have abi_enable, but sysvipc_enable,
> > linux_enable and svr4_enable.
>
> I looked at this. abi is in fact a container for three start scripts.
>
> Of course splitting them makes the situation worse..
>
> If the ${name}_enable check becomes a function
> some special cases could be treated in a case statement
>
> case ${name} of
> mountd)
> if [ -f /etc/exports ]; then result=1; fi
> ;;
> abi)
> if [ checkyesno sysvipc -o checkyesno linux_enable .. ]; then
> result=1; fi
> ;;
> *)
> enable=eval \$${name}_enable
> if [ "X${enable}" = "X" ]; then result=1
> else result=`checkyesno ${enable}`; fi
> ;;
> esac
>
> [That's untested demo code and may contain typos]
>
> Unfortunately the rc.d scripts are not self-contained anymore. That may
> offend some. But there is shared *.subr code already.
>
the initial idea behind rcng (was it called like that?), was, and most of it
is still good, it's nicer that the monolithick /etc/rc of old, and less messy
than
sys5's. lets try and keep it 'simple' - as usual the KISS principle has
more than one interpretation :-)
> > All these - and some more that I probably missed - can be fixed, or the
> > warnings ignored, but is it worthwhile?
>
> That's a good question. It is up to the users of "slower" hardware
> (e.g.embedded devices), I think.
>
yup, on the other hand, servers take so long to boot nowadays, that speeding up
rc, IMHO, wont make a difference, or even if it's fast, a fsck that's running
in
the background will slow the host anyways :-).
> Regards
> Peter
cheers,
danny
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?E1KKaRd-0008h9-Hl>