Paul,

Thanks for your note.  See comments inline below.

Steve


On Tue, Nov 11, 2025 at 1:07 PM Paul Wouters <[email protected]> wrote:

> On Tue, 11 Nov 2025, Paul Hoffman wrote:
>
> [mostly responding to Steve's message]
>
> >> 1. Transitioning to hyperlocal root zone access, as including designing
> and deploying the infrastructure to distribute copies of the root zone to,
> say, a million hyperlocal caches.
> >
> > The DNSOP WG would be the right place to specify how to do "hyperlocal
> root zone access", as it did for RFC 7706 and RFC 8806. It is unclear where
> work on transitioning to that mechanism would go. Maybe the IAB (it's
> architecture), maybe DNSOP (it's DNS operations), maybe a new WG (it's a
> departure, not an extension, to the current DNS), ... .
>
> Didn't it already do everything? It is now up to the OSes to use the
> protocols we defined?
>

As more and more resolvers switch to using local copies of the root zone,
there may be a need to strengthen the process of providing those copies to
a very large set of resolvers.  That's the part that is not yet scoped and
may need design and implementation.

>
> >> 2. Redesign of the root zone update process to be fully encased in a
> tamperproof enclosure, with updates of each portion of the zone requiring
> cryptographic approval by the relevant TLD operator.  A key ceremony
> equivalent would also exist for the exceptional cases to override the
> regular process, eg initiate a new TLD, replace lost device, etc.
> >
> > Maybe the IAB (it's architecture), maybe DNSOP (it's DNS operations),
> maybe a new WG (it's a departure, not an extension, to the current DNS),
> ... .
>
> I am not sure what this adds to the current deployment. The TLDs give
> DS'es to the root. If there is ever a signature of root over something
> the TLD operator did not submit as DS RRset, or a deep signed record
> within that TLD for something else (SVCB, ECH, HTTPS, TLSA) then you
> would have cryptographic evidence of root misbehaviour.
>
> If you want something that can prevent (some) abuse, you would be
> better of having the independent rootzone operators provide a DNSKEY
> and RRSIG over everyhing and run the root with 13 keys :P And you can
> do "n of m" or just trust the root operators you trust for whatever
> personal reasons you have, for example just trust F :)
>

The primary purpose of such a design is to prevent improper forceful
removal of legitimate entries from the root zone.  This is stronger than
detecting the problem after the fact.



>
> Paul W
>


-- 
Sent by a Verified

sender
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to