[ntp:questions] Re: NTP Multicast server/client?

Danny Mayer mayer at gis.net
Sun Oct 2 15:11:49 UTC 2005


David Woolley wrote:
> In article <433E0D98.6070004 at gis.net>, mayer at gis.net (Danny Mayer) wrote:
> 
> 
>>Wilhelm Greiner wrote:
>>
>>>Also i see when ntpd with debugging runs that it transmits any packets
>>>to find out is the way back reachable.
>>>
>>
>>I don't understand what you mean here. Can you explain?
>>
>>
>>>Can i turn off this initial trials??
>>>
>>
>>I also don't understand this question.
> 
> 
> When a broadcast client is starting it makes a normal access to the
> server to calibrate the round trip delay and therefore estimate the
> downstream delay.  It will fall back to the configured delay if this
> fails.
> 

I'm not sure that it does send a packet if authentication is not 
enabled, but I'd have to check.

> Also, I believe that if you enable authentication, and that is strongly
> encouraged for broadcast, this exchange is mandatory to negotiate keys.
> (I vaguely remember that authentication is defaulted on for broadcast.)
> 

Authentication is enabled by default for both broadcast and multicast. 
Turning it off means that you have accepted the risks of doing so.

When you authenticate that means that you are exchanging a bunch of NTP 
packets between broadcast/multicast client and server. It's described as 
the "key dance" in the documentation. I don't remember how many packets 
are exchanged, but it's more than two.

Danny



More information about the questions mailing list