Re: [Ntp] Antw: Re: Follow-up to yesterday's mic comment about PTP security

[Miroslav Lichvar <[e-mail]>]

On Thu, Jul 25, 2019 at 09:06:30AM +0200, [e-mail] wrote:
> None of that matters for what Daniel is saying.
> A packet can still not have arrived *faster* than with lightspeed via the 
> shortest line between the two devices.
> And that's all you need to derive that the situation is in fact marginally 
> better than "half of RTT (plus oscillator-related delta)".

There is a variant that can be used in local networks with known
HW/topology. Instead of physical distance, which is negligible in
local networks, the metric is the number of network devices between
the stratum 1 server and client. If the minimum forwarding delay of
all devices is x (under all conditions) and packets between the server
and client always have to go over at least n devices, the root
delay reported by the client can be reduced by 2*n*x. The maximum
error of the client's clock (root distance) then may be shown to be
smaller than the required accuracy (e.g. 100 microseconds in the case
of MiFID).

With PTP (which assumes a constant network delay and doesn't have a
concept of root delay/dispersion) it's more difficult to determine the
maximum error.

-- 
Miroslav Lichvar