In message <[email protected]>, at
09:42:18 on Sat, 25 Jan 2014, Jim Reid <[email protected]> writes
It might be true that the majority of registrants just stick with
whatever DNS is offered by their registrar but not all of them do that.
Clueful ones certainly don't.
For some value of "clueful".
I expect the majority of registrants don't worry very much about
continuity of service, nor would they even notice if their website was
offline as a result (and the number who use something where an
interruption might be more noticeable, like domain-based email rather
than various cloud and connectivity-ISP-based email, must be an even
smaller minority).
Their "clue" is more of a financial sort, where they are happy to pay a
few tens of dollars a year for the less resilient service, compared to
something much more expensive for the greater resilience.
That's partly why I said, earlier, that "Best practice is supposed to be
that they should be separated, although many commercial hosting
companies appear not to."
Then there's the issue of hosting organisations who apparently put two
Name Servers in the same /24 [for our non-technical readers that's two
servers on the same branch-of-a-network-with-254-usable-IP-addresses,
previously called a Class C; something that typically has no
connectivity redundancy, even if such a design could cope with one of
the two servers failing].
This is all a subset of a general theory which states that "when
Internet users became so numerous that someone gave up trying to publish
an annual list of them all in a paperback book, lots of stuff changed".
{Was it 1994 - I have that book, bought in 1995...
http://www.amazon.co.uk/The-Internet-White-Pages-1994/dp/1568843003 }
The best thing we as a WG can do is try to acknowledge that such changes
*have* happened, that we have 2 Billion users, and when we are giving
advice to Governments and Regulators it should be appropriate for a
World with 2 billion users, not the 100 thousand trusted users that many
clearly wish it still was. That boat sailed in 1995.
--
Roland Perry
