(Fwd) Top 100 Maintainers List

Hi Niall,

Let me make some remarks about this. We probably should this dicuss
in details this later at the meeting.

    This message from the RIPE-NCC Database Group shows that a
    number of domain registries appear among the top-scoring
    maintainers of inconsistent objects in the RIPE Database.

Yeah, and we're the winner! And we could have done better :-). I'll
discuss the situation further on.

    Irrespective of the future of the TLD-WG, we need to identify
    how best to address these concerns and, in particular, this
    problem of inconsistent objects.

Let me start to raise some questions. Citating (probably incorrectly)
Daniel Karrenberg, ``Ripe is in the IP-number business, not in the
naming business''. Given that, one wonders whether domein name
objects should be part of the Ripe database at all. Since Centr is
now in de name business, I wonder whether control of this part of
the database should not be adressed bij Centr.

I also wonder what the added value is to have Ripe mirror the whois
database for a ccTLD. If the ccTLD has a whois database it is the
only authorative one. A pointer in the ripe database could be
enough.
-------------

The NL-DOMREG has apparently 51932 inconsistencies.

Maybe this is a somewhat too detailed story, but on the other hand,
it is a nice case-history of how these things go.

Since I'm quite new in this position, I researched this a bit and
apparently the next of events took place.

In the beginning of RIPE's existence the NL-DOMREG feed, together
with the SE-DOMREG feed, to Ripe's database was one of the better
functioning ones. The form in which the objects were--and I believe,
still are--send was slightly adapted so only relative global
information was inserted in the Ripe database. For critical
information such as admin-c and tech-c info was a pointer to the
NL-DOMREG whois database. This was apparently done to prevent the
data in the ripe database becoming stale. Since the update was/is
done every workday, the chance of this hapening isn't big, but
still more then zero.  Another reason was protection of the objects.
For a while was it possible that arbitrary persons, not necessarily
part of the NL registry, could change .nl domain object. I've heard,
that now this shouldn't be possible anymore.

This went well for a while until Ripe changed the policies concerning
the contents specification of the objects. The NL registry has
asked for an adaption of these rules or the way the Ripe database
can be manipulated. Since that time no updates have been done.

There was also another problem. We have access controls on our
whois database to limit the amount or completely block searches in
the database to prevent data mining for commercial purposes. This
didn't align with the Ripe policies at that time.

Since that time no updates have been done for the Ripe database.
Today we just handed out our 100.000nd domein registration, so
about 50% of the .nl objects are not found in the ripe database at
all.  We are currently seeing around 7000 new domains each month,
so this number is growing rapidly.

The inconsistencies themselves are mainly caused by the old pseudo
person ``see remarks'', which was/is the pointer to the nl regsitry
whois server.

Also, some joker apparently grabbed the ``not maintained'' object
and changed it, causing for more disaster.

Currently we are implementing a complete Domain Name Registration.
With the design we have included the new Ripe update policies and
we expect that we will refresh and update all the information as
soon as this system comes in production. Regular updates will then
resume.
-----------------------

This is as far as I understand what happened with the .nl objects
in the ripe database consulting various persons. I might have some
details wrong.

	jaap



-------- Logged at  Fri Sep 10 09:47:18 CEST 1999  ---------