[ntpwg] [dhcwg] NTP option: IP address and/or FQDN

[Richard Gayraud (rgayraud)]

Hello Danny,

I think you are right. This was discussed during both the NTP working 
group and the DHCP working group. At some point, it was told by some
people that the address it better and the worries in NTP community are
due to misunderstanding of how DHCP works. But I believe this is not 
true.

Their main argument is that DHCP is not configured statically, and has
the ability to resolve DNS names, and will, in turn, redistribute this
as an IP.

  => This ignores the NTP pool capability

  => But more importantly, this does not give any guaranty that a vendor 
     will not ship a small home router with a DHCP server inside, with an 
     embeded NTP server address => worst case.

This is why I think we really need to address FQDN. I had the opportunity 
to discuss with a few people after the meetings, and had the feeling this
can be accepted as a consensus.

The syntax to use for this is less consusual. It could be :

  - an option with a 'type-switch' byte : people do not like this
    solution.

  - an option with sub-options (one sub-option for address or one for
    FQDN, plus a few optional ones for parameters like Key-ID)... 
    (most flexible solution)

  - 2 different options. They would have the same additional parameters, 
    which is not that nice...

Again, each option can appear several times in the DHCP message,
so that several NTP serves can be configured.

One question for you Danny : In ntpd.conf, I think we have to 
configure pools with a different keyword than 'servers'. Does this 
means that a pool must be differentiated in the config ? Should it 
be carried as another sub-option in DHCP ?

Thanks,

Richard. 

> -----Original Message-----
> From: Danny Mayer [mailto:mayer at ntp.org] 
> Sent: dimanche 9 décembre 2007 05:19
> To: MORAND Lionel RD-CORE-ISS
> Cc: DHC WG; NTP Working Group; Ralph Droms (rdroms)
> Subject: Re: [dhcwg] NTP option: IP address and/or FQDN
> 
> MORAND Lionel RD-CORE-ISS wrote:
> > Hi all,
> > 
> > During the meeting, we discussed again the use of FQDN for 
> advertizing
> > NTP servers.
> > 
> > During the work on the PANA option, we had exactly the same 
> discussion
> > (in the meeting and on the mailing list) and the conclusions was the
> > same: we SHOULD use IP address list instead of FQDN. The 
> SHOULD is there
> > to say that it might be some cases where the use of FQDN would be
> > valuable/really required.
> 
> Don't you think that you should have also discussed this with the NTP
> folks as well? Most of the DHCP discussion only happened 
> because we were
> asked to review the draft.
> 
> > As it seems that, after at least two long discussions in 
> the same topic,
> > there is a "common" understanding on how should be 
> advertize servers, it
> > could be useful to capture these conclusions as guidelines in the
> > dhc-option-guidelines draft. This would help a lot other 
> working groups
> > when they will have to design new options for the same usage.
> 
> I'm not yet convinced that this conclusion is really warranted since
> this will force NTP to use only a single IP address for each server
> instead of being able to use the pool option which provides a lot of
> flexibility. It's not in the NTPv4 draft but that shouldn't 
> matter since
>  this is about provisioning NTP and not about a specific 
> implementation.
> Don't forget that NTP will have to live with the decisions you make
> about what you tell it to use. It's either that or the 
> options will get
> ignored and neither result is desired. I'd like to discuss 
> how this will
> be operationally implemented before we can make a decision on 
> whether or
> not IP addresses are the way to go.
> 
> I am copying the NTP WG on this so we can get other people's input.
> 
> Danny
> 
> _______________________________________________
> dhcwg mailing list
> dhcwg at ietf.org
> https://www1.ietf.org/mailman/listinfo/dhcwg
>