KPNQwest ns.eu.net server.

Tue Jun 11 18:17:17 UTC 2002

On Thu, 6 Jun 2002, John Payne wrote:

> I found it interesting to note that a significant number of cctld servers
> ignore the suggestions for root-servers in BCP40/RFC2870...
> "Other major zone server operators (gTLDs, ccTLDs, major zones) may also find
> it useful." and leave recursion enabled on the ccTLD servers (2.5) - the old
> ns.eu.net was one of these, I believe RIPE have done the right thing with the
> new one.

A lot of the older secondary nameservers for ccTLDs were also the
recursive nameservers for the ISP/Organisation providing the secondary
service.  ns.eu.net is a classic example of this.

With the valid quips about how long it takes to update glue/NS sets in the
roots[1], a fair number of these ISPs/Organisations had found that
shifting the ccTLD secondary function to a proper non-recursive server[2]
was simply not practical.

--==--
Bruce.

[1] teckla.apnic.net, trf.nic.ad.jp, etc[3]
[2] Some ISPs do still 'need' to allow recursion to cater for their
    roaming customers.  imo, customers are easier to change than the root.
[3] some quick stats on the hosts mentioned in the root ('.') zone from a
    viewpoint in Amsterdam:

    Number of records:                   657
    Number of fully valid hosts:         481
    Number of partially valid hosts:     110
    Number of invalid hosts:             175
    Number with reverse matching:        455
    Number knowing about themselves:     551
    Number not knowing about themselves: 106

    fully valid = all of the nameservers for the domain the nameserver is
	in know about the nameserver (all NS for example.com answer for
	for ns1.example.com)

    partially valid = some of the NS for the domain the nameserver is in
	do not know about the nameserver in question.  Note that the
	answer is skewed slightly due to multiple answers received.

    invalid hosts = This host only exists in the root glue.  No
	nameservers for the domain the nameserver is in know about the
	nameserver.  Answer possibly skewed due to my assumption in what
	the 'parent' domain for the nameserver is (cut -d '.' -f 2-)

    reverse match = name -> A -> PTR == name

    knowing about self = Asking the ip in the root glue for the name gives
	a sensible answer.  (imo, this is a Good Thing, but unfortunately
	I don't believe that any exact requirement for this exists)