[ntpwg] SNTP and NTP servers

[TS Glassey]

Dave - I want to change the license for NTP so that people are not allowed
to use it for actions like what DLink did to various people on this list who
are still fielding its 'sales support' overhead.

It is exactly the NTP License that has caused all of the 'Bad 
Implementation' issues IMO. No one should be allowed a license to implement 
a 'portion' of the NTP service because it prevents the rest of us from being 
safe around it.

----- Original Message ----- 
From: "David L. Mills" <mills at udel.edu>
Cc: "NTP Working Group" <ntpwg at lists.ntp.isc.org>; <dhcwg at ietf.org>
Sent: Monday, December 03, 2007 9:53 PM
Subject: Re: [ntpwg] SNTP and NTP servers


> Danny,
>
> The NTP specification defines primary servers, secondary servers and
> clients. Primary servers are synchronized to reference clocks and
> conform to the SNTP or NTP specification, which is the same.

IMHO - this is a problem not a solution. The SNTP handshaking/protocol
should be constrainable so that NTP Servers can refuse service to SNTP
Client's... The reason that Wisconsin happened is that a Commercial Entitiy
violated the generally accepted rules. The point is that the reason that was
possible is that the original license allowed them to. If the license had
been properly written, then anyone would have 'recourse' and what folks like
DLink did to Poul-Henning and myself would have been avoidable.

> SNTP
> clients can synchronize to primary or secondary servers, but cannot
> offer synchronization to clients of their own.

I.e They can recieve synchronizations from NTP Servers,  but cannot in turn
offer that service to other client's. I.e. SNTP is limited to be used as a
receptor of time data and not a transmitter of it.

> Secondary servers can
> import time from primary or secondary servers and provide time to
> dependent clients. Only they must provide the mitigation and discipline
> algorithms.

OK. My only pushback here is that there is NO mandatory implementation model
based on the "License" here - so we cannot control what is implemented and
what is not.

> This is to avoid possible whiplash or other suboptimal
> behavior. If you have alternate discriptions or verbage about the
> specification prose, I would be happy to hear it.
>
> Dave
>
> Danny Mayer wrote:
>
>> Just to clarify, SNTP clients can use both NTP Servers and SNTP Servers
>> while SNTP servers cannot be used for NTP clients unless the have a
>> refclock attached. Yes this a bit confusing but it is described in the
>> current NTPv4 draft. See Section 14 of the draft. The SNTP packets are
>> the same as the NTP packets.
>>
>> Danny
>> _______________________________________________
>> ntpwg mailing list
>> ntpwg at lists.ntp.org
>> https://lists.ntp.org/mailman/listinfo/ntpwg
>
>
> _______________________________________________
> ntpwg mailing list
> ntpwg at lists.ntp.org
> https://lists.ntp.org/mailman/listinfo/ntpwg