Re: [Ntp] Of Roughtime's algorithm agility, and host attestation

[Watson Ladd <[e-mail]>]

On Fri, Jul 26, 2019 at 11:32 PM Patrik Fältström <[e-mail]> wrote:
>
> On 27 Jul 2019, at 7:58, Watson Ladd wrote:
>
> > It seems people are misunderstanding the trust relationship here. The goal is to have publically observable, auditable performance and identities tied to that. So it makes very little sense to link these identities to another naming system, but maybe this is a better idea then it seems. In particular public keys really shouldn't change: that could permit the laundering of errors, which we very much don't want.
>
> Ahhh...ok, sorry. I did not write enough words.
>
> I agree with the base idea in what you write here. One should be careful on what one trust. But, one also need to base the trust on something, some root.
>
> This can of course as some have suggested a cert for some time service one trust. Or the root CA of something else. Like root in some other naming system.

I consider roughtime analogous to certificate transparency: the point
is not that Nimbus or Behind the Sofa are necessarily run by good
people or that the certs are logged by them, but that for the past X
they have dutifully logged each one, and if they were to fail to,
there would be incontestable proof of malfeasance. Nothing is really
rooted in the assertion of identity which is sort of irrelevant here.

>
> My answer was "just" responding to the question whether we should in the case some cert is stored in DNS create a new RRType, which is something I find not necessary as TLSA exists.
>
>    Patrik



-- 
"Man is born free, but everywhere he is in chains".
--Rousseau.