Subject: Re: sysctl(2) and/or /kern for system variable manipulation
To: NetBSD Kernel Technical Discussion List <>
From: Greg A. Woods <>
List: tech-kern
Date: 03/25/2000 23:18:31
[ On Saturday, March 25, 2000 at 13:35:39 (-0800), Greywolf wrote: ]
> Subject: Re: sysctl(2) and/or /kern for system variable manipulation
> Greg.  Go run Plan 9 if that's what you're after. :-)

Plan 9 and Research UNIX are not the only systems that have demonstrated
with real-world experience the benefits of unifying the system (and
network) namespaces into the filesystem.

> I disagree with everything living in filesystem namespace.  It's an added
> level of interpretation, context and complexity that doesn't need to be
> there.

In fact the overall level of complexity, from pretty well all points of
view so far as I can tell, is actually lower.

It is indeed an extra level of interpretation (i.e. as opposed to
looking in /dev/kmem at the actual raw data structures).  However it is
this level of interpretation that allows the interface defined for
external programs to be a little more resilient to minor changes in the
internal details and indeed even to some extent to be machine

Indeed even the 4.4BSD developers were well on their way to
demonstrating the benefits of a unified filesystem namespace for all
system objects in their borrowing of things like procfs from Research
UNIX and the development of portals, etc.  I don't doubt that they would
have gone further in this direction had their group continuted to work
on BSD.  Note that everything that McKusick et al say about the
advantages of sysctl apply equally well to an interface designed as a
virtual filesystem (and they go almost so far as to say this too).

							Greg A. Woods

+1 416 218-0098      VE3TCP      <>      <robohack!woods>
Planix, Inc. <>; Secrets of the Weird <>