On Mon, Sep 24, 2007 at 10:58:19PM +0000, Pierre Habouzit wrote:
> On Mon, Sep 24, 2007 at 10:08:17PM +0000, Sergio Talens-Oliag wrote:
> > El Mon, Sep 24, 2007 at 08:54:38PM +0200, Pierre Habouzit va escriure:
> > > On Mon, Sep 24, 2007 at 06:08:35PM +0000, Sergio Talens-Oliag wrote:
> > I accept that my patch is not good enough for your taste, and maybe a better
> > one or a different solution is a better option, but I disagree about your
> > argument about conffiles. 
> 
>   It's not a matter of "not good enough" it's just brittle. My job is to
> provide a good sane default for 99% of the use of the software. I can't
> provide a perfect sane default for any use, so just make up your stuff
> if you need to.

  I may not be very clear, but here is what I think:
  * either you provide a rock solid way to know which interface the user
    listen to, and sed is not good enough ;
  * or you roll your own configuration.

  Any in between state is somehow more than unsatisfying, and is a
situation that will likely generate bad configurations. I reckon the
current situation isn't great for what you try to achieve, but you don't
make it better. Make it better means, prevent users to shoot themselves
in the foot while configuring. Using resolvconf _and_ listening to
anything else than any or the loopback is just unlikely enough that I
don't want to provide a quite broken support for this.

-- 
·O·  Pierre Habouzit
··O                                                [EMAIL PROTECTED]
OOO                                                http://www.madism.org

Attachment: pgpV0zANmwv7K.pgp
Description: PGP signature

Reply via email to