Re: [Ntp] Splitting the Roughtime draft?

[Marcus Dansarie <[e-mail]>]

On 2021-01-31 09:39, Marcus Dansarie wrote:
> I think it's important to at least define a basic serialization schema
> for both malfeasance reports and lists of servers. If we don't, we risk
> ending up with several incompatible formats for data exchange in the
> Roughtime ecosystem.
> 
> A de facto standard for server lists already exists (ecosystem.json) and
> only needs to be documented in the draft. A simple malfeasance report
> format in the same spirit as the server list format could be a JSON
> object with an ordered list of objects, each containing a nonce and a
> server response, both in base64 format. Verifying the report would be as
> simple as (1) verifying that the list is a valid list of chained
> Roughtime responses and (2) applying the same validity tests as clients.
> 
> Tal Mizrahi and others raised a number of issues with how the security
> model is described in the draft. [1] I don't think they where ever
> addressed. We should probably be proactive in fixing this in the draft,
> as it is something that the IESG will have opinions about. In
> particular, I think we need to be very clear about the fact that all
> trust in Roughtime is rooted in the long-term keys and that they are
> expected to be valid for a very long time indeed.
> 
> After fixing these two two points, I would be happy to support a WGLC.

I just realized that I forgot to add one important point: We need to
specify how revocation of long-term keys is performed, even if it is as
simple as knowingly generating invalid time responses and impeaching
oneself.

Kind regards,
Marcus