[Resolvconf-devel] Bug#339418: Bug#462946: Bug#462946: resolvconf reverses nameserver orderand then truncates early second time if 127 is

[Peter T. Breuer]

"Also sprach Thomas Hood:"
> Peter T. Breuer wrote:
> > Actually, that looks completely incomprehensible
> 
> 
> You have to unzip the file first.
> 

:-). No, it's not zipped here on debian. It has been compressed with
gzip, but of course that makes no difference to my ability to read it!

> If you read the file unzipped then I can only conclude that you cannot

??

> understand resolvconf and therefore you are probably better off without

I haven't read it at all, apart from the HOWTO section.  Does it say
something more intelligible in the rest than the following examples,
chosen at random ...

 webmin-wvdial (8)
     Copies /etc/ppp/resolv.conf over /etc/resolv.conf after PPP
     connection established unless (as of version 1.160-3) the
     latter is a symbolic link; restores original contents from backup
     file when the connection is broken.
     
 xisp (1 -- removed) Adds lines
     to /etc/resolv.conf on PPP interface up and deletes those lines on
     interface down.

which aren't even english sentences.

I'll repeat the question - I don't know why you are getting narked
about it! Where in this file does it tell me how to do what we ALL want
to do with resolvconf, namely set a new set of entries in resolv.conf
when an interface is changed to a new IP address?

Can you point to that please? 

As to the HOWTO, the only bit that looks likely to be relevant is

    3.4 ifupdown
        * Remove any "up" or "down" commands from
        * /etc/network/interfaces that futz with /etc/resolv.conf and
          remove any scripts from

I'm afraid I can't do the latter, since that's debian's business. But
the former I can do, and there aren't any (the only such entries are for
using resolvconf to set a given list of nameservers, which it doesn't
do any longer since the last upgrade, thanks to the 127 business).

The bit about

   3.6 bind
       * Change the /etc/bind/named.conf file so that it includes
         /var/run/bind/named.options instead of
         /etc/bind/named.conf.options.  The former will be generated
         from the latter, as needed, by inserting or updating the
         "forwarders" statement inside the öptions" statement with a
         current list of forwarders.

Is probably quite helpful, but irrelevant to me, since I don't want
resolvconf to control the local named - that's none of your business,
in principle.  It's just a caching server for some entries known only to
me.  It should remain untouched.

Incidentally, named.conf does not do any includes here. Mind you, I
like the way you have generated that file

   /var/run/bind/named.options

and I WILL think about including it. But I'm afraid I don't want to
forward to the forwarders. It's just a caching server in case of
emergencies, when the forwarders don't work, and for private domains.
So there would be no point in doing the including (I've thought :).
It would even be counter-productive.

Still, producing that file to be included is cute!

As far as I can tell, that's where the HOWTO ends.

Now, as to your design objectives of opaqueness, you are meeting that
criteria, but you're second guessing what I want when an interface
changes, and doing so wrongly. I don't want the list of nameservers to
be short-circuited.

> it.  As root do:
> 
>     dpkg --purge resolvconf

Why? Isn't it part of debian? How would debian work otherwise? As far
as I can see, it's pretty basic.

> Please follow up at #339418.  This means that next time you should write
> to 339418 at bugs.debian.org instead of to #462946.

I'm just pressing "reply"! If you want to set a different reply-to
address, please see the manual for your mail user agent, whatever it is
:). Or possibly the README, compressed or not.

I'll repeat my observation - when 127.* is in the nameserver list
passed to resolvconf, resolvconf has a BUG. Not only does it truncate
the list at the point where 127.* is found, but it truncates it BEFORE
that. I don't know if you intended that or not. Maybe, maybe not. 
But if you would kindly just truncate at the point where 127.* appears,
then I could get it to do what I want, instead of what you want, which
is different from what I want. Second guessing the user just makes the
user give up on you ... you want to give the user more control. That's the
aim of a tool, no?

Peter