Tracy R Reed wrote:
I have been a big fan of distributed hash tables (DHT's) for the
efficient distribution of information for a few years now. Ever since I
first heard about Freenet, really. Freenet itself has given up on much
of the DHT concept for various reasons, most having to do with
anonymity. Kademlia is another very interesting DHT. Bittorrent uses the
Kademlia concept to good effect. But there are much more important
infrastructure type issues that could make use of DHT's to dramatically
improve the reliability and efficiency of the Internet. DNS is one place
where a DHT would seem to be a very good solution:
http://www.cs.cornell.edu/people/egs/beehive/codons.php
Here is a nice summary of how it all works which I found while googling:
http://www.cs.toronto.edu/syslab/courses/csc2231/05au/reviews/HTML/16/0001.html
The problem is that this is using a hammer to swat a fly.
What DNS really needs is for major ISP's to get regular zone transfers
with incremental update from the roots. A good example of this
hierarchical structure is ntp.
An ntp client is never supposed to hit a Tier-1 or Tier-2, it is
supposed to hit something much more local (see the fiasco with Netgear
and Poul-Henning Kamp for what happens when this breaks down).
Generally, every ISP runs a Tier-3 time service. This should be the
same with DNS.
Then, recursive queries stop at the ISP 99.9% of the time. In addition,
the DNS system can quite easily survive the loss of a root since the
ISP's would just hold the entire table until the next zone transfer
occurs (ie. no timeout at full zone level).
All it would take is one single root and one major ISP to start making
this work. Then network effects can take over.
-a
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
