Date: Thu, 11 Sep 2003 00:30:24 -0700 From: Terry Lambert <tlambert2@mindspring.com> To: Doug Barton <DougB@FreeBSD.org> Cc: Philipp Grau <phgrau@ZEDAT.FU-Berlin.DE> Subject: Re: new rc system Message-ID: <3F602490.CDD98F0D@mindspring.com> References: <20030908105911.GA152151@CIS.FU-Berlin.DE> <20030910164901.T35663@12-234-22-23.pyvrag.nggov.pbz>
next in thread | previous in thread | raw e-mail | index | archive | help
Doug Barton wrote:
> ntpd[ate] is a very difficult thing to order, because a lot of things
> need/want accurate time before they start, and yet by definition, ntp is
> a network protocol so it has a lot of other dependencies before it can
> even start. A lot of the things that ntp depends on (like network) want
> logging before they start, but syslog wants accurate time before IT
> starts... and we spiral back in time infinitely.
Technically, syslog only has a soft dependency on the current time,
since the time is looked up each event, rather than being cached.
So it has a dependency, but it's not one that prevent syslog from
running or providing some minimal/degraded service level.
> > This is a known shortcoming in the new rc system. Luke Mewburn
> > commented on it in a talk recently but does not yet have a
> > satisfactory solution.
>
> Can you describe in more detail what you mean by "this is a known
> shortcoming?"
I believe he means the distinction between "soft" ("would like")
vs. "hard" ("absolutely cannot run if not present") dependencies.
The new rc system fails to make such distinctions: it doesn't
have a high enough granularity to its dependency classification.
-- Terry
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3F602490.CDD98F0D>