On Tue, May 31, 2005 at 08:08:41PM -0700, Gordon Waidhofer wrote:
> 
> I concur with Bill's proposal of a standardized ioctl()
> for firwmare downloads.
> 
> Having device nodes (ultimately with sufficient access
> control) seems like good discipline and not wasteful
> at all. I've always viewed having the name as part of
> the ifconf ioctl as a kludge. It is inconsistent with
> so many other device operations (disks, ttys, etc).
> Further, it precludes any access control model other
> than suser().

What he said.  It seems to me that the decision long ago
to _not_ give network devices names in the filesystem namespace
was a bad one.

As I've asked before, how many namespaces do we really need to
have in our system (beyond system calls and filesystem names,
I suppose) and why, exactly, is it a good thing for those
namespaces to proliferate?

This is something that Plan 9 got very, very right.

Thor