On Mon, May 18, 1998 at 11:20:32PM +0200, Miquel van Smoorenburg wrote: > In article <[🔎] 19980518221139.F13509@tapiola.infodrom.north.de>, > Martin Schulze <joey@infodrom.north.de> wrote: > >On Mon, May 18, 1998 at 09:36:12PM +0200, Miquel van Smoorenburg wrote: > >> No, as I said there's not enough space in K90-K99. But the other option > >> is not too bad - we could indeed extend the runlevels to allow numbers > >> from 100-999 as well, which would run after 0-99. It would not be strictly > >> alphabetically anymore though. Or maybe allow hex numbers? KA0 - KAF or > >> so.. that would keep the alphabetical order in tact. > > > >Sorry, I object. Could you explain that more detailed? Why 9 > >levels aren't enough to catch the scripts? > > Well, we already have K91apache, so only K92 - K99 are available. It's > almost impossible to fix this on existing running systems you know. We > also have to assume a user has not used K92 - K99 him/herself (which > we can't) or otherwise the system will break. K91apache is a mistake in the bo version, it should be K10apache. Recent versions of apache use S91 and K10. So please ignore it. So the highes levels are: lrwxrwxrwx 1 root root 13 Mar 14 11:50 /etc/rc0.d/K89atd -> ../init.d/atd lrwxrwxrwx 1 root root 18 Mar 14 04:46 /etc/rc0.d/K90sysklogd -> ../init.d/sysklogd These could be easily changed to K79 for atd and K80 for sysklogd to have another 10 free numbers. The sysklogd maintainer is around and I we have permission from the at maintainer to do NMU's if needed, iirc. > We already use 5 K levels, and people may want to add more. We only have > 8 K levels free if we asume they are not used by the local sysadm. > > What if we choose to use > > K92sendsigs > K93urandom > K94umountfs > K95mdutils.sh > K99halt > When I want to insert a "foobar" script between sendsigs and urandom, that's > no problem. I can use K93foobar. But at that moment K93 has been used twice, > something you want to avoid. Because now package "argh" comes along and you > want to put it between K93foobar and K93urandom .. impossible unless you > change the name. We could move them a little bit more into the 8x region. > That's why I said it's better to use sparse allocation of the Kxx links. > Because however unlikely this scheme sounds, Murphy's law guarantees a > situation like this will surface. Sure. Regards, Joey -- / Martin Schulze * joey@infodrom.north.de * 26129 Oldenburg / / A mathematician is a machine / / for converting coffee into theorems /
Attachment:
pgpGdeqjzbaH5.pgp
Description: PGP signature