[ntpwg] ***SPAM*** Re: [dhcwg] Network Time Protocol (NTP) Server Option for DHCPv6, WG item updated

[Danny Mayer]

Jyrki Soini wrote:
>> The option you propose falls into this category. Moreover, it looks 
>> like it is not supported by current implementations of NTP (does not 
>> exists in the reference implementation config file). This makes me 
>> think it is not really required on the the NTP side.
>>
> 
> I think it is not that much an NTP issue, but multicast usage issue.
> Source specific multicast is somewhat simpler and more robust to 
> implement on network infrastructure. Please refer mbone and magma 
> mailing list archives.

Not for NTP. NTP's main concern is that the server can be authenticated. 
not so much where the multicast packet supposedly came from. Source 
addresses are easy to spoof. Having the right keys to get authenticated 
using autokey is much harder to do and is the preferred way of accepting 
multicast packets. There are also different ways that we use multicast 
packets to handle isolated networks for example.

Danny