On Thu, Feb 25, 2016 at 10:34 AM Paul Hoffman <[email protected]> wrote:

> On 25 Feb 2016, at 10:18, Ted Lemon wrote:
>
> > I'm sorry to be a sticky wicket here, but I have to ask: have you
> > thought about what a guaranteed-correct implementation of this would
> > look like?   I think you need to actually do that analysis before we
> > proceed with this.
>
> Can you say more? It seems like the spec in the draft is a
> guaranteed-correct implementation: if you have a current validated
> statement that nothing exists between N1 and N2, and you later get a
> query for something between N1 and N2, send back NXDOMAIN directly.
>

As discovered by Geoff Huston, Google Public DNS (8.8.8.8) already does
this[0].


>
> > As best I understand it, getting this right is not trivial,and

> getting it wrong would be harmful.


I think that the "getting it right" for a DNSSEC validator is fairly
trivial[1], but I fully agree that cocking it up would be harmful.

W

[0]: IIRC, this was while collecting data for his "On Queries to the Root"
presentation.  Sorry for not saying this earlier, I wanted to clear it with
folk / not step on toes.
[1]: much more so than many of the other thingies that DNSSEC validators do!



> While it clearly would help in
> > the context of widespread adoption of DNSSEC, I'm not convinced that
> > the security risk of the added complexity would be compensated for by
> > an actual reduction in woe at the root.
>
> Please say more about the "security risk". I'm missing it.
>
> --Paul Hoffman
>
_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop

Reply via email to