Re: [dnsext] we need help to make names the same, was draft-yao-dnsext-identical-resolution-02 comment

[Phillip Hallam-Baker <[e-mail]>]

On Wed, Feb 23, 2011 at 5:25 PM, Donald Eastlake <[e-mail]> wrote:

> Hi,
>
> On Wed, Feb 23, 2011 at 12:00 PM, Phillip Hallam-Baker <[e-mail]>
> wrote:
> >
> >
> > On Wed, Feb 23, 2011 at 5:30 AM, Tony Finch <[e-mail]> wrote:
> >>
> >> On Tue, 22 Feb 2011, Phillip Hallam-Baker wrote:
> >>
> >> > If you are going to do [online signing], you might as well do a key
> >> > exchange inline as well as we do in TLS. One key exchange can then be
> >> > leveraged across multiple connections using kerberos style tickets
> (see
> >> > DPLS for an example).
> >>
> >> That gives you channel security whereas DNSSEC gives you data origin
> >> authentication. They are not the same things.
> >
> > True, but data origin authentication is probably the wrong model for a
> DNS
> > security scheme.
>
> Why? Is your goal to make it easy for some entity not the authority
> for a zone to forge data in that zone?


Data origin authentication is very expensive. If the use cases only require
channel authentication, that is going to be more practical.



> > If we are going to consider changing the model of DNSSEC, which is what
> > moving to online signatures would entail, then the whole architecture is
> > back on the table.
>
> Total nonsense. The on or off line signing question is pretty minor
> and completely orthogonal to the channel versus origin authentication
> question. As soon as you have a zone with dynamic update, your are
> shoved in the direction of on-line signing, it doesn't take mixed case
> non-ascii.


The original poster was being imprecise. DNSSEC already has an 'online'
signature model in that the only practical way to sign a zone is to have the
signing keys connected to the Internet.

What is being proposed here is really an inline signature scheme in which
the signatures are generated within the DNS request-response loop. It would
require every DNS server to have signature capability, not just the zone
master. It would require the signatures to be generated on demand in
response to attacker generated queries.

It would be a total change of the security model.


I think that it would be unwise to make such a change to the DNS
infrastructure in order to support a requirement that is probably nonsense.


-- 
Website: http://hallambaker.com/