Re: CVS commit: src/sys/dev/pci

[David Holland <dholland+sourcechanges%eecs.harvard.edu@localhost>]

On Fri, Feb 29, 2008 at 06:46:46PM +0100, Joerg Sonnenberger wrote:
 > > right, I don't see how this is related to environmental sensors / power 
 > > management.
 > 
 > Let's follow FreeBSD's decision and drop the whole environmental sensor
 > framework in terms of custom sysctl nodes all over the tree.
 > 
 > Seriously, RAID state is a environmental factor and making it as easy
 > accessible as possible is a good thing. It is not like this is adding a
 > lot of code...

It is only "environmental" by stretching the definition of
"environmental" a long way.

I don't think it's bad to have this logic, but I think it needs to be
framed and presented in a way that makes sense, and that means things
ought to appear under the right names and in sensible places.

Really, if you were looking for ways to set up RAID monitoring and
failure notification, would you look in /etc/*powerd*?

Ultimately we need a general framework for responding to things that
happen in kernel, whether those things are low-battery notices, RAID
failures, device insertions, or whatever. But there are quite a few
unanswered questions about how that framework ought to work, and I
don't think it's a good idea to ignore those questions and stuff in ad
hoc mechanisms that later become the de facto answers.

(I am also not thrilled about the proliferation of scripts-as-config-
files. One of the major unanswered questions is how responses to
events should be configured. "Just run a script" is expedient but poor
design.)

-- 
   - David A. Holland / dholland+netbsd%eecs.harvard.edu@localhost