[Ntp] Antw: [EXT] Re: The bump, or why NTP v5 must specify impulse response

>>> "Dieter Sibold" <dsibold.ietf@gmail.com> schrieb am 16.04.2020 um 19:27 in
Nachricht
<17906_1587058115_5E9895C3_17906_233_1_E591F38A-A303-4077-BC4C-A61AB7867F83@gmai
.com>:
> The problem with this combination is that we currently have no 
> implementation that conforms to RFC5905 completely.
> There are different use cases for potential clients. Some of them are 
> not well served with the current algorithm other are. A modular approach 
> would allow both: to apply an algorithm most suitable to the application 
> and to be compliant with the protocol spec.

Hi!

In the past I tried to read "version" to find out what the implementation is
conforming to. One remarkable exception was Solaris NTP that suppressed the
version (obviously for "security" reasons). The other alternative was to guess
from the variable list which version the remote end was. Then (again for
"security" reasons) most implementations shut down the control protocol too.
Maybe NTPv5 should require some more verbose version field. Also considering
the few bits the version field has right now, we need a better mechanism for
the future anyway, I guess...

One of the amazing facts of NTPv4 implementation was that variable names as
presented in the control protocol changed names. Like these:
CS_DRIFT,     RO, "frequency" (once "drift")
CS_ERROR,     RO, "clk_jitter" (once "error")
CS_STABIL,    RO, "clk_wander" (once "stability")

I think a future specification should require fixed names for specific
variables, too Or maybe assign numeric labels like four digits.
So if the semantics of a variable changes, it would get a new number...

Regards,
Ulrich

> 
> Dieter
> 
> 
> On 16 Apr 2020, at 11:58, Harlan Stenn wrote:
> 
>> On 4/15/2020 6:38 AM, Salz, Rich wrote:
>>>>    We had the same discussion at the transition from ntpv3 to ntpv4.
>>>>    If you do this, please do not call it NTP.
>>>
>>> Can you explain why 3‑>4 could still be called NTP but 4‑>5 should 
>>> not?
>>
>> You didn't quote enough context.
>>
>> NTP is a combination of two parts:
>>
>> ‑ the protocol/packet structure
>> ‑ algorithmic and response behavior
>>
>> If you are only specifying the protocol in this new thing then it is
>> dangerous/misleading/irresponsible to call this v5 document "NTP".
>>
>> It is only slightly less dangerous/misleading/irresponsible to call it
>> "NTPv5 Protocol", regardless of whether or not there is also an "NTPv5
>> Algorithms and Response Behavior".
>>
>> ‑‑ 
>> Harlan Stenn <stenn@nwtime.org>
>> http://networktimefoundation.org ‑ be a member!
> 
> _______________________________________________
> ntp mailing list
> ntp@ietf.org 
> https://www.ietf.org/mailman/listinfo/ntp 