Re: RIPE Region Hostcount for September 2000

[Lee Wilmot <]

Hi,

Daniel Rosen wrote...

>> The de domain shows the highest increase for September with
>> approximately 35,000 hosts more than in August.
>
>Shouldn't that be .nl?
>
>nl       320304     1947310      529824     1417486    + 54471

Yes, apologies, human-error strikes again. I live in Holland,
I should have noticed this :-)
  
 * > Is this loss due to a technical problem or some general chang
 * es
 * > (named-based vhosting instead of IP-based vhosting)? Or just
 * > "fluctuation"?
 
Every time I've investigated large drops, they've turned out to be due
to one of two things. Either...

- an ISP which runs secondary for it's customers from a single
DNS server restricted access for one reason or another, thus
removing lots of litte-medium size zones from the count all at once

or

- an ISP or end-user with a large zone restricts AXFR on all it's
servers. In this case, the number of hosts for a TLD can drop
dramatically, even though the number of zones counted still increases.

Simon Leinen wrote...

>(2) by moving to a "split" DNS scheme, in which only hosts with
>    external services (in the "DMZ") are listed in the DNS, the
>    internal hosts being in another version of the organisation's zone
>    which is only visible from the inside.
>
>The latter is proably difficult to work around.

The hostcount, despite the name, does not count hosts, it counts A
records. Increases in the use of split-DNS, name-based virtual hosting
etc have meant that the correspondence between the number of A records
and the number of hosts has rapidly decreased. There's not a lot that
can be done about this within the framework of the hostcount.

However, there are several things that can be done to combat the
AXFR-block spread. For instance, when asked, most organsations (ISP's
and end-users both) will open up transfers to the collecting machine
once it's clear what the purpose of the AXFR is. Unfortunately, most
of the resources available for the hostcount have been applied to
running the hostcount every month, not much has been left for
maintenance and other issues.

>Asking for such cooperation would be easier if there was a strict
>privacy policy for the intermediate data produced by the hostcount
>software.

We have an AUP for the hostcount data on the RIPE NCC website and
only grant access to the data based on signing of this AUP.

   http://www.ripe.net/hostcount/aup.html

Unfortunately, there's a hole in this policy because most local
collectors still make the hostcount data publicly available.
Coordination of the closure of this hole is part of the maintenance
activities of the hostcount for which no time has been available.

Regards,

Lee Wilmot
RIPE NCC