Re: [Ntp] Suggested improvements to the Roughtime draft (draft-ietf-ntp-roughtime-00)

[Marcus Dansarie <[e-mail]>]

On 2020-02-17 21:23, Tony Finch wrote:
> The TF.460 requirements are largely about radio signals for celestial
> navigation, so it's questionable that those aspects are relevant for
> distributing civil time over computer networks. TF.460 also only requires
> DUT1 to 100ms precision, so where does your 1ms precision requirement come
> from? GPS provides DUT1 to sub-microsecond precision, for example.
> (IS-GPS-200 table 30-VII, https://www.gps.gov/technical/icwg/)

When I wrote the suggestion, I was envisioning distributing the DUT1 as
announced by IERS, which is done with a precision of 0.1 s. The
millisecond precision in my suggested format was selected mostly
arbitrarily. Since there are sources that distribute DUT1 with much
higher precision, I would not be opposed to changing the resolution to a
more suitable one. This would also add the need for a way to indicate
the DUT1 resolution.


> There is no need to distribute both DTAI and a leap second list. (DTAI can
> be derived from the list.) I think a new time protocol should just
> transmit the entire list, because it can easily be compresed to less than
> 20 bytes - the current list is 16 bytes and we're unlikely to pass 20
> bytes within the next 10 years.

I view DTAI as (somewhat) separate from the leap seconds list, in that
it works even if the server does not have any information about past or
future leap seconds. For example, the GPS signals provide the current
UTC-GPS offset.

Using the LEAP tag, the server can send all past leap seconds if it
wishes. Since all Roughtime data comes in four byte increments, I would
strongly prefer any format in the LEAP field to be compatible with this.
I am, however, open to better data representations that the one I suggested.


Kind regards,
Marcus