Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 2 Jun 2009 10:48:33 -0500
From:      Brooks Davis <brooks@freebsd.org>
To:        Doug Barton <dougb@dougbarton.us>
Cc:        freebsd-rc@freebsd.org
Subject:   Re: Removal of deprecation for network_interfaces != AUTO
Message-ID:  <20090602154833.GE14685@lor.one-eyed-alien.net>
In-Reply-To: <4A2456DA.8040104@dougbarton.us>
References:  <4A21A4F6.5060709@dougbarton.us> <20090601212506.GA2351@lor.one-eyed-alien.net> <4A2456DA.8040104@dougbarton.us>

next in thread | previous in thread | raw e-mail | index | archive | help

--u5E4XgoOPWr4PD9E
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Mon, Jun 01, 2009 at 03:31:54PM -0700, Doug Barton wrote:
> Brooks Davis wrote:
> > On Sat, May 30, 2009 at 02:28:22PM -0700, Doug Barton wrote:
> >> Without objection I plan to commit the attached patch before the code
> >> slush, and to MFC the change.
> >=20
> > I object.  Supporting values other than AUTO adds unnecessary
> > complexity (not a lot, but some) and IMO leads to difficulty diagnosing
> > proper system behavior.=20
>=20
> Can you provide examples? I've seen this argument before, but I
> seriously fail to understand it. The code to generate the list for
> AUTO is trivial, and I've never seen a support question that was
> caused by setting this to a different value.

It's been a while since I worked on the code and I'm busy so I'll
have to back off on that argument for the moment.  I do remember it
complicating the logic, but don't have the time right now to recall what
I wanted to do.

I will assert that the second argument is wrong.  The deprication notice
has reduced the number of errors related to network_interfaces a fair
bit.  The biggest problem before was that people would omit lo0, but I
seem to recall there were several other misunderstanding that removed
feet.

> >> I've never seen the rationale for this, and I use a value other than
> >> AUTO personally for a script I have that tests to see if the wired
> >> interface is up and starts the wireless if not. I've also seen other
> >> users ask about this from time to time, so I'm sure I'm not alone.
> >=20
> > Please provide this script to support your argument.
>=20
> I set the value of network_interfaces to "lo0 test" and then I have a
> /etc/start_if.test script that checks to see if my wired interface is
> up, initializes it if it is, and if not it checks to see which
> wireless card I'm using and initializes that. I've actually spent
> quite a lot of time trying to figure out how to accomplish something
> similar with the current system but AFAICS we don't have a way to do
> that without hooking it in at the point where the network interfaces
> are actually configured.

Wouldn't /etc/start_if.<wired> do the same thing without having to set
network_interfaces?

> At some point in the future when I get a whole bunch of free time I'd
> like to extend what's in the base in order to do what I'm doing now
> with my little script ...

I'd agree we could use some better hooks for this sort of thing.

-- Brooks

--u5E4XgoOPWr4PD9E
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (FreeBSD)

iD8DBQFKJUnRXY6L6fI4GtQRAv14AJsG2kjPbIZd6fiVJY/UumjVdDjZygCfWXew
zk/iSFX9WAO9X886HfgCX28=
=0Moo
-----END PGP SIGNATURE-----

--u5E4XgoOPWr4PD9E--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20090602154833.GE14685>