[ih] Funny how things work out
karl at cavebear.com
Mon Feb 1 19:09:51 PST 2021
On 2/1/21 7:23 AM, Dave Crocker via Internet-history wrote:
> The pre-ICANN effort of the IAHC included a survey of various DNS
> experts. While there was no consensus about the technical limits of
> the DNS or its operation, there was a pretty solid view that a couple
> of thousand TLDs would not be a technical problem.
Peter Deutsch (remember Archie, one of, if not the first search engine?)
and I decided to try to see if we could make a root server go boom by
stuffing it with a mountain of TLDs.
This was done, of course, entirely off line.
What we did was to grab a copy of the entire .com zone at the time (it
was much smaller then, but still many millions of names) and "elevated"
it to be the root zone in a Bind based server on a reasonably well
endowed PC of that era (small by today's standards.)
We then wrote some code that threw a synthetic query stream at the box,
with a controllable mix of invalid queries and other things to try to
annoy its internal caching. We set up several of these synthetic
traffic generators, lit the fuse, and watched.
It did not collapse, it didn't even really break out in much of a
sweat. We were somewhat astonished.
Our conclusion what that even way back then that there was a lot, and I
mean a lot, of headroom in the root zone for expansion. And that was a
couple of decades ago.
My sense was that the root size limits were not so much on the number of
TLD entries but more on the areas of propagation of updates and the rate
of human error maintaining all of those entries.
Later on I was part of a crew that filled out and submitted the better
part of a hundred new TLD applications to ICANN. Whew! (But at least we
got to spend those months mostly in some extraordinarily nice
locations.) [There's a whole interesting side story about how we and
other teams were able to game ICANN's "digital archery" ;-) ]
I much agree with the notion that many top level domains are not very
I'm also of the belief that DNS is slowly fading out of the eyes of
users and becoming more a part of the internal machinery of the net
rather than something we ought to fight about. See
Some years back I looked at naming issues in cloud computing and asked
whether DNS was adequate. My conclusion was that DNS forms a really
solid, rock solid, foundation for relatively non-dynamic name-to-record
mapping, but that cloud computing presented some issues that are beyond DNS.
These included the need to deal with entities in "the cloud" that could
partition and rejoin. I wrote a note about that - On Entity Associations
In A Cloud Network -
I am much of the belief that we can learn a lot from biology. In
particular, I was struck by many users of the net who are want to locate
an instance of a thing rather than the, singular thing. For example, if
I am looking for a station to charge my car I am interested in a thing
with characteristics of "supercharger" and "near me" rather than search
for some particular instance.
Living entities have a similar need to find things, like mates, based on
attributes (e.g. my species, opposite sex, near me) rather than a given
It seems to me that the Internet could use something on top of DNS that
is more like a search engine than a name-to-record mapping engine.
Web search engines seem too much attuned to human users than
programmatic ones, and more tuned to keywords than some open set of
meta-attributes. There was a protocol some time back, IF-MAP, that was
going down that road; I don't know what happened to it, but it felt like
it was pointing in a needed direction.
(For years, sort of as a joke, sort of seriously, have proposed the
notion of Internet pheromones as a means of locating resources. But I
always run into the same wall - a protein or other complex chemical in a
pheromone can carry far more information that we can stuff into any
rational size data packet; and chemistry seems much better suited to
fast pattern matching than most of our computers.)
(I could also go off in the direction of how we could adopt some methods
from astrophysics into the net but I've gone far enough astray already.)
More information about the Internet-history