[dns-operations] The (very) uneven distribution of DNS root servers on the Internet

[Warren Kumari]

On May 17, 2012, at 12:25 PM, David Conrad wrote:

> Warren,
> 
> On May 17, 2012, at 9:20 AM, Warren Kumari wrote:
>> On May 17, 2012, at 11:57 AM, David Conrad wrote:
>>> On May 17, 2012, at 3:25 AM, Joe Abley wrote:
>>>>> Even ignoring folks who slave the zone now, is coordinated measurement of the root system realistically possible today given the business/political/philosophical environments of the root operators?
>>>> Yes.
>>> 
>>> [citation needed]
>> 
>> DITL
> 
>> CAIDA
>> TLDMon
>> DNS Measurements with RIPE Atlas Data
>> DNS-OARC
> 
> Do any of these cover all instances of all root servers?

Not as far as I know, but they cover enough to give you a picture of what's happening…

I took a Statistics 101 class in university. First day the lecturer provided an example scenario -- you run a rubber band factory and want to to figure out how strong your rubber bands  are. You can be completely sure of your result by testing each and every band till it breaks, and then calculating an average… but then you don't have any rubber bands to sell -- so obviously you should just test a (large enough) sample…. I say "obviously", but for some reason the class just *couldn't* wrap their brains around this. After the third session discussing this I got bored and wandered off…

I don't believe that you can (nor need to) cover all instances of all root servers. Folk are able to resolve example.com (and dancing hamsters.net). This works, and works well. There are folk who care deeply about this, and are, IMO, doing a fine job of it (and I've read most of the reports / easements from stuff like the above, chatted with many of them, etc., so I ain't just tooting' smoke). 

I realize others have different opinions, but at the moment I've gotten bored with this discussion, and am wandering off…

Regards, 
W
> 
> Regards,
> -drc
>