On 2022-04-19 23:33:06 +0300, Michael Tokarev wrote:
> On Wed, 5 Jan 2022 16:36:40 +0100 Vincent Lefevre <vinc...@vinc17.net> wrote:
> ..
> > But I don't understand. The upstream nameservers are supposed to be
> > used as a fallback. Even if upstream nameservers do not perform DNSSEC
> > validation, this is still better than a failure when DNSSEC is not
> > required.
> 
> For the record, this is incorrect, just like has been stated in #1004032
> numerous times already.
> 
> The upstream nameservers provided by DHCP were never supposed to be used
> as a "fallback", even more, there's no _notion_ of a "fallback" in this
> context.
> 
> We EITHER use the DHCP-provided nameservers, OR we use the regular recursive
> way.  But not both.
> 
> I know no recursive resolver software which has notion of "fallback" like
> this.

Without resolvconf installed, it appears to work: if unbound cannot
resolv the hostname, then the next nameserver in /etc/resolv.conf is
used.

For instance, I currently have in /etc/resolv.conf:

nameserver 127.0.0.1
nameserver 192.168.1.1

If I stop unbound, then I still get hostname resolution. But if I only
have

nameserver 127.0.0.1

and unbound is stopped, then hostname resolution no longer works.
This shows that 192.168.1.1 is used as a fallback.

And something like "strace wget ... |& grep sin_addr=inet_addr" also
confirms this behavior.

-- 
Vincent Lefèvre <vinc...@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)

Reply via email to