Re: [DNSOP] [Fwd: New Version Notification for draft-vandijk-dnsop-ds-digest-verbatim-00.txt]

[Brian Dickson <[e-mail]>]

On Fri, Sep 25, 2020 at 5:36 AM Peter van Dijk <[e-mail]>
wrote:

> Hello dnsop,
>
> in this new episode of 'enabling future innovations that we perhaps
> cannot even imagine today', please find below a link to a draft
> proposing a DS digest type that does no digesting at all. This allows a
> zone owner to publish information in the parent zone and have the
> parent sign that data on the owner's behalf.
>

No.

Please read all of rfc5507, and in particular section 5, on why this goes
against IAB guidance,
years of operational and developer experience, and is a horrible, horrible
idea.

Do not overload (re-use) RRTYPEs, period.

Also, beyond the potential use case of having parent-side non-authoritative
data DNSSEC signed,
by using a different RRTYPE for the delegation side of a zone cut, and
having a way of discriminating
glue records (such as A and AAAA) from non-glue equivalents, so that the
parental size of a zone
cut can DNSSEC sign these records, the entire design of the Domain Name
System is built around
the delegation principal, which is that the only place data is
authoritative is below a zone cut.

Having the ability to push data up to the parent side, which is controlled
by the child side, violates
the implied security model of the relationships between zones, which
roughly correspond to the
relationship in the Unix file system of directories, symbolic links, and
mount points. The corresponding
mechanism would be to allow an unprivileged user to "give away" ownership
via the "chown" command,
to a privileged account. This capability was extensively abused in the Unix
world until it was blocked.

Implementing this mechanism in DNS would almost certainly be abused in
similar ways, with the same
end result, of having to disable that mechanism due to abuse. This is
reason enough not to even start
down this road.

I'm happy to provide example use cases, but don't want to distract from the
general model being bad,
by playing whack-a-mole on examples and retorts. (See Warren Kumari's email
signature line about
pants and weasels for why.)

With respect and no offense intended.

Brian