Re: [ietf-provreg] References for Today's Host Object Discussion

[Jaap Akkerhuis <jaap@sidn.nl>]

Joe,
    
    There is one related issue that may deserve some thought. At the moment 
    an address is required for a host record if the host is named under the 
    apex of the registry. Where hosts are specified as attributes of domain 
    objects, however, there is the opportunity to modify that rule; 
    addresses can be required on hosts which are named under the domain in 
    whose object they appear.
    
    So, for example, a domain object for "automagic.org" might contain host 
    attributes for nameservers "buffoon.automagic.org", "ns1.flame.org" and 
    "foo.something-else.com". Only the "buffoon.automagic.org" host would 
    require addresses under the scheme outlined above.
    
    In the case of a registry which supports hosts-as-attributes, this (or 
    something like it) seems like it would be necessary to avoid different 
    domain objects containing conflicting addresses for hosts.
    
    
Well, as a registry which does that, we actually also like to have
ip#s for out of zone name servers (optionally). We check wether the
delegation exists for the domain at the mentioned nameservers.

If no ip#s are given, we use whatever we find from the DNS. If the
ip#s are given, we use those. This turns out to be useful when more
then one things are happening at the same time. The ip#s of name
servers itself might be changing as well (because of a merger of
parties involved etc.).  In cases like these the proper A record
might not be available yet.

Of course, we never publish the A for "foo.something-else.com" and
the like. We do keep the information archived. We keep an archive
of all transactions surrounding domain names. Just in case
(technical/juridical) problems pop up later.


	jaap