Re: [Ntp] NTS Pools

[David Venhoek <[e-mail]>]

Dear Miroslav,

I have added the row with computational cost, and for reference also
included a few other interesting computational cost parameters. As to
regard to your new proposal, I was wondering how that would work with
the key derivation? I don't see a way to get the key derivation on the
TLS session with the backend to result in exactly the same key as that
coming from the session with the client (in fact, if it could I'm
wondering whether that would be a security issue for TLS). It is
exactly this trouble that leads to the server changes in the NTS-KE
load balancer case.

Kind regards,
David Venhoek


On Wed, Feb 28, 2024 at 11:33 AM Miroslav Lichvar <[e-mail]> wrote:
>
> On Wed, Feb 28, 2024 at 09:29:47AM +0100, David Venhoek wrote:
> > Hi All,
> >
> > I have made a spreadsheet showing a short overview of the implications
> > of the various options, see
> > https://docs.google.com/spreadsheets/d/1rxEArrbN5OKgsFI9KUzmOzI1twDztQc6aEUZr5WelVU/edit?usp=sharing
>
> This is great. Thanks!
>
> Can you please add a row comparing computational cost of TLS wrt
> number of clients for the pool provider (e.g. 0, 1, N, 2N)?
>
> I have another proposal to consider. The pool provider could simply
> terminate TLS and forward NTS-KE communication between clients and
> individual NTS-KE servers run by the NTP providers. No modifications
> required on clients or servers. The NTS-KE servers would just need to
> be configured to provide their IP address in the server negotiation
> record, so the client doesn't send requests to the pool's TLS host.
> The TLS cost for the pool provider would be 2N.
>
> --
> Miroslav Lichvar
>