Re: Nameserver MUST HAVE IP

[Peter Eisenhauer <eisenhauer@schlund.de>]

On Mon, Apr 09, 2001 at 08:03:30AM -0400, Hollenbeck, Scott wrote:
> >-----Original Message-----
> >From: Mats Dufberg [mailto:dufberg@nic-se.se]
> >Sent: Monday, April 09, 2001 7:26 AM
> >To: Hollenbeck, Scott
> >Cc: ietf-provreg@cafax.se
> >Subject: RE: Nameserver MUST HAVE IP
> >
> >
> >On Mon, 9 Apr 2001, Hollenbeck, Scott wrote:
> >
> >> >IMHO we should remove the requirement in 3.4.2[3] that a 
> >nameserver MUST
> >> have
> >> >an IP address if its name lies registered in the 
> >registrar's TLD. The
> >> >discussions here and the interesting report quoted by 
> >Elisabeth Porteneuve
> >> >showed that this requirement does not add much to the "health"
> >> >of the domain name system.
> >>
> >> I disagree that the requirement should be removed 
> >completely, and reiterate
> >> my earlier suggestion that replacing the MUST with a SHOULD 
> >would be more
> >> appropriate.
> >
> >A SHOULD would be appropriate if storing the IP address for glue records
> >not strictly needed would be a better solution than not. I disagree. By
> >storing and announcing the IP address you increase the risk of have
> >different data in the mother zone than in the authoritative 
> >zone. -- Data should be kept where it is authoritative as far as possible.
> 
> ...and by not mentioning anything at all about the need for an IP address
> with a name server you run the larger risk of not having glue where it's
> needed.  I don't have a problem with adding some words that describe why the
> SHOULD isn't a MUST (see below), but cutting out the requirement completely
> doesn't address the reality of needing addresses to produce glue records.
> 
> 3.4.2-[3] existing text:
> Name servers registered within the registry's authoritative TLDs MUST be
> registered with a valid IPv4 or IPv6 address.
> 
> 3.4.2-[3], change MUST to SHOULD and add new text:
> Name servers registered within the registry's authoritative TLDs SHOULD be
> registered with a valid IPv4 or IPv6 address only as required to produce DNS
> "glue" records.
> 
> In my opinion this ensures that we still have a requirement for addresses to
> produce glue records, and covered the concern that addresses aren't required
> for all name servers.
> 
> <Scott/>
> 
> <Scott/>

From my point of view this seems to be a good solution.

Best Regards,
Peter Eisenhauer

-- 
Peter Eisenhauer                        Schlund + Partner AG
Erbprinzenstrasse 4-12                  Tel: 0721/91374-0
76133 Karlsruhe                         E-Mail: eisenhauer@schlund.de