[dns-operations] TLD(s) for private use

[James Stevens]

> Nope. If someone in that setting wants to have globally unique names, they need to do so by registering a globally unique domain name. If there's no requirement in these setups for globally unique names, they don't need to do that. For instance they could rely on unmanaged BonJour-style naming services.

I'm sure you are right, but my experience is that this is often not the 
case - i.e. /most/ DSL routers give out a domain name to the LAN, which 
is usually just some random word - i.e. TLD squatting.

Mine hands out "lan" and that can't be changed in the WebUI :)


It would be interesting to hear from a ROOT server operator (or two), or 
somebody who has looked at the OARC DITL data, to know what proportion 
of ROOT query traffic might be the result of TLD squatting.

If there was a recognised collection of "unregistered" domains, those 
queries could be caught at the ISP.


> BTW, my home LAN does not have a DHCP server or rfc1918 addresses.

I'm sure you are aware I was talking about the general case and I'm sure 
you are aware your situation is very unusual.

However, it /appears/ that Andrew, despite knowing that applying for 
some registered IPv4 address space for use at home is the "correct" 
solution, has not done so, but instead appears to tolerate issues that 
arise from rfc1918 use.

> On 06/09/17 16:35, Andrew Sullivan wrote:
>> Oh, if_only_  all network engineers understood such implications!  If
>> that were true, then my printer would not occasionally turn out to be
>> on the network of Oracle's London office whenever I connected to the
>> VPN.  1918 address space is in fact problematic because it is supposed
>> to have_only_  local significance, but people don't use it that way
>> because of NAT.

Like I said, rfc1918 issues, especially collision, is as much an issue 
of general lack of IPv4 address space as anything else.


James