Re: dhc* in NetBSD 7.0

[Thor Lancelot Simon <tls%panix.com@localhost>]

On Tue, Mar 26, 2013 at 02:02:23PM -0400, Christos Zoulas wrote:
> On Mar 26,  1:29pm, tls%panix.com@localhost (Thor Lancelot Simon) wrote:
> -- Subject: Re: dhc* in NetBSD 7.0
> 
> | On Tue, Mar 26, 2013 at 03:22:33PM +0000, Christos Zoulas wrote:
> | >
> | >   fix it. I think that it is appropriate to include dhclient in the
> | >   next formal release, and announce that this is the last release where
> | >   we will ship it, to give a chance to people to upgrade.
> | 
> | But this was the exact same response I got to my suggestion that we
> | drop dhclient for the 6.0 release.  How do we know we won't go around
> | on this again when it's time for NetBSD 8.0?
> 
> We forgot to announce it's demise in the 6.0 release notes. Let's not repeat
> this. In fact, let's start the release notes for 7.0 now, so that we can
> add more things we want to deprecate.

Let's announce the demise of dhclient in the release notes for 6.1 or 6.2,
then.  I don't see any reason why .0 releases should be special for this
kind of thing.

> | We have plenty of time before NetBSD 7.0 ships to ensure dhcpcd works
> | for all reasonable use cases.  Its maintainer is a NetBSD developer and
> | has made his intention to ensure this is so very clear.  NetBSD 7.0 is
> | the right time for dhclient to die.
> 
> Except for the cases we did not get a chance to test. Right now dhclient
> is a minor 4+MB annoyance in /sbin, not worth fussing too much about it.

I don't see a 4MB executable as a "minor" annoyance.  I see it as one of
the larger single opportunities to get duplicative bloat out of the base
system.

Roy has committed to fix any bugs about which anyone actually gives him
useful bug reports.  He has done an excellent job of that in the past,
and implemented a considerable amount of backwards compatibility (for
example, the hook scripts) to make life easier for people with custom
or complex dhclient configurations.  I can't see how it's sensible to wait
years more before dropping code we could reasonably have dropped in 6.0.

Thor