> > As a Node.js maintainer I also ran into this problem. Our solution was
> > to ignore these fec0:: servers entirely
>
(https://github.com/piscisaureus/cares/commit/f6cdd7d404d44eb2d6e8866e85ae65
1b89fc769e),
>> and we've never had any complaints about it. I
> > think pycares did the same. Placing these fake DNS servers at the end
> > might be a good solution as well, I don't really have a strong opinion
> > on that. However I would prefer to try the "real" servers in the order
> > that windows reports them.
>
> But these entries are there for some kind of purpose, right? Won't there
be
> systems that actually rely on their presense in that list to function
properly?
Windows implements a draft RFC from 2001
(http://tools.ietf.org/id/draft-ietf-ipngwg-dns-discovery-03.txt) that
suggests that DNS resolvers try a couple of fixed "site-local" IPv6
addresses (fec0:0:0:ffff::1, fec0:0:0:ffff::2 and fec0:0:0:ffff::3), which
would allow the networking stack to find DNS servers without any
configuration.
I have never encountered a network where a DNS server was actually available
at one of these addresses.
The whole concept of site-local addresses is now considered problematic and
the use of the fec0::/10 address range has been officially deprecated in RFC
3879 (http://www.ietf.org/rfc/rfc3879.txt) so it seems unlikely that this
way of setting up DNS on a network will become popular in the future.
> > Lastly I think it'd be good to filter the list of DNS servers and
> > remove duplicate entries. If there's consensus that this is a good
> > idea I can put together a patch that does this.
>
> I can't see any benefit from having duplicates, no. Is this common?
Windows associates DNS server addresses with specific interfaces (weird, I
know). When there are multiple network interfaces present (including virtual
ones) it'll walk all of them and return all DNS servers configured for this
interface, without filtering out duplicates. So yes, it's somewhat common.
- Bert
Received on 2013-04-23