[discuss] Interesting article

Roland Perry roland at internetpolicyagency.com
Tue Jan 14 19:03:58 UTC 2014


In message <1389718795.70149.YahooMailNeo at web121002.mail.ne1.yahoo.com>, 
at 08:59:55 on Tue, 14 Jan 2014, nathalie coupet 
<nathaliecoupet at yahoo.com> writes
>I would really like to understand the technical requirements for this 
>continued connectivity. Would they use CGN tools or plug-ins?

I'm going to use .uk and Nominet (the UK ccTLD registry) as an example, 
because it simplifies the explanation. Thus the ISP, and its customers 
in this example, will be in the UK.

First of all, connectivity is largely based upon IP addresses (rather 
than DNS). Customers of the ISP are allocated IP addresses for the 
purpose of connecting to their ISP (either public IP addresses in the 
classic case, or private IP addresses in the case of carrier-grade NAT).
So that leg of the connectivity is independent of the DNS.

If the ISP itself is using .net or .com domain names for its 
infrastructure, then its customers, and people contacting its customers, 
are entirely insulated from fun-and-games happening in '.uk'

So we now have just those ISPs who have placed their own infrastructure 
within .uk, and therefore most likely their customers have been assigned 
.uk based names.

All that the root server system does (and it is frequently over-sold) is 
tell someone asking, where to find Nominet's servers; which is a first 
step towards finding out where someone (ISP or their customer) with a 
.uk domain name is located in IP Address space.

The IP address of Nominet's servers is not something which changes very 
often (a timescale measured in years I'm sure).

So all our example ISP has to do to keep its own users in contact with 
other potentially isolated .uk domains is to make sure that the DNS 
server they provide for their users (and which is inherently under the 
control of ISP) continues to give that rarely-changing answer about 
Nominet's servers.

After that, the user's DNS enquiry will inevitably (although out of 
sight of most users) proceed normally, using the data provided first by 
Nominet, and then by the name servers nominated by the .uk registrant.

Meanwhile, in the rest of the world (including other bits of the UK), if 
the operators of DNS servers refrain from expiring the DNS entries 
pointing at Nominet then inwards connectivity-by-name to .uk sites will 
be preserved.

----

But that's all in the hours or days after this hypothetical deletion 
from the file sent to root servers. You can be sure that in the weeks 
and months after that, measures would be put in place to ensure long 
term continuity irrespective of whatever the root server operators 
(several of whom are for-some-value-of independent) are "instructed" to 
put in their published copy-of-the-root.

----

I'm aware that some readers will regard the above as either 
over-simplified, or over-complicated. I'm very happy to respond to 
either (on or off list), in order to refine the explanation of the 
process.

>________________________________
> From: Roland Perry <roland at internetpolicyagency.com>
>To: discuss at 1net.org
>Sent: Tuesday, January 14, 2014 11:29 AM
Subject: Re: [discuss] Interesting article


In message <1389715417.54174.YahooMailNeo at web121001.mail.ne1.yahoo.com>,
at 08:03:37 on Tue, 14 Jan 2014, nathalie coupet
<nathaliecoupet at yahoo.com> writes
>Could you explain how removing a country from the Internet in the root
>zone would not prevent this country from being connected to the 
Internet?


Because .com isn't a country?

Because even .uk was removed, all that would disappear (modulo caching
and out-of-band methods of keeping the IP address of the servers in
question in circulation) to a first approximation is .co.uk websites and
email servers. It wouldn't even affect users in the UK's ability to look
at websites in, or send email to .fr, let alone .com. And they'd get
replies to their emails as long as they were using an email address
outside of .co.uk

[nb there are other .uk sceond level domains, omitted for clarity].

-- 
Roland Perry



More information about the discuss mailing list