On Mon, 13 Oct 2014, Phillip Hallam-Baker wrote:

DNSCurve was originally proposed to do a rather different job and it
is actually better at that job than DNSSEC in my opinion.

I agree that DNSSEC sucks at the job it is not supposed to do :)

I would expect that anyone taking up DNSCurve and proposing it in this
forum would do what I did and use it as a starting point and make
modifications.

You want different spokes to jam in the dns caches? Or a different spoke
to stab into the authoritative DNS servers' CPU? Or a different spoke to
stab into the DNS proxies and deep packet firewall inspectors and DNS
transparent proxies?

There are very good reasons dnscurve was not picked up by the IETF. And
it really doesn't make sense to pick up dnscurve for only the transport
hop encryption part, because:

1) it is hacked into the DNS using NS record overloading with severe
   limitations
2) it spews non-DNS traffic over port 53
3) it hardcodes crypto parameters to the One True Curve
4) it was designed to run on authoritative servers, not dns caches

Really, dnscurve was about killing the dns caching resolver. Why on
earth would anyone propose to use it to secure those same entities the
protocol insists on killing?

dnscurve has no place in an IETF dnspriv discussion other than to serve
as an example on how to not secure resolver-stub communication.

I'm not set for any of the proposed solutions yet, but:

- I like the idea of using (D)TLS (but on port 443, possibly with ALPN,
  and not using any kind of "starttls hack" to put non-DNS traffic on
  port 53)

- I like the idea of using regular format DNS payloads over port 53
  but nervous about more problematic bad DNS software

Paul

_______________________________________________
dns-privacy mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dns-privacy

Reply via email to