[dns-operations] Karl Auerbach on adding 'millions' more TLD - what do folks think about the operational impact?

[Florian Weimer]

* Stephane Bortzmeyer:

>> If there are millions of TLDs, it's no longer so easy to run a root
>> service instance.
>
> You mean that today, it is easy? :-)

I haven't tried, but I was under the impression that running yet
another local node for F root, say, is not that hard from a technical
POV.  Depending on the location, query rates can be far below what
other name servers see.  Keeping the zone in sync shouldn't be a
problem, either, given it's glacial change rate.

Things change a bit when zone load times etc. become an issue once
there are 2,500,000 records instead of 2,500.

> Many people on this list manage domains with > 1 million
> delegations. I do not think they would call their job "easy" but it
> is not science-fiction.

I don doubt that.  Especially in the lower 7 digit range, you can run
a server without near-real-time updates on commodity hardware
nowadays, I suppose (and every three or four years, we gain another
digit or so).

Real-time updates are another story.  Could you run them on the
current root server infrastructure?  (I don't think so.)

> So, yes, basically, Auerbach is right: the current limitation of the
> number of TLD is not motivated by technical concerns (we are very far
> from the practical limits).

And to some degree, you can control how fast the number of TLDs grows,
even if the delegation requirements are lowered significantly.

However, other things become even more difficult than they are today.
If you've got an issue with bar.foo, do you need to contact foo, or
bar.foo, and if the latter, how do you get some contact point from
foo?