iCNS is a DNS-like name service that is part of the iCite net. iCNS gives names to content, and so, on the Internet, functions a layer above DNS.

In creating a new system like this, I have been trying to understand if there are liabilities in DNS that don't need to be built into iCNS. And, the main issue seems to be the concept of "root servers".

I had a conversation with Nick Levay about this last week at the Joi Ito dinner, and he didn't see the removal of root servers offering a lot of technical advantages. The main advantages that many people seem to see with eliminating root servers are political / social, as the role of ICANN and VeriSign in the DNS system are often considered harmful.

So, I can understand how a system based on root servers can be corrupted if the root servers go down or are themselves corrupted. In the iCite net, the iCite Containers (which are the main clients of iCNS) mostly work peer-to-peer, but take advantage of the authoritative iCNS for some look-ups (e.g., for discovering new iCites and for forwarding when iCites move to new locations) and for double-checking information they get from their peers.

What I could do is design the iCite Containers to "know" more about how to work without iCNS, but to take advantage of iCNS as long as it is there. With DNS, this might be like if we had an option to bypass VeriSign if we wanted to.

But, I don't yet have a good picture of how things would be better with a purely p2p system. (I first mentioned this last month in my post on Overlay / peer-to-peer networks).

I am reading up some good materials about p2p DNS right now, and would like to hear / read more opinions on this topic. Here is what I am studying now:

YURL: Secure Referencing, and especially the white paper, Naming vs. Pointing.

Names: Decentralized, Secure, Human-Meaningful: Choose Two, from Bryce "Zooko" Wilcox-O'Hearn, the lead on Mnet.

I am also looking at Java JXTA resources, including DisDNS and p2psockets.

Suggestions and comments appreciated!