Regarding the PeoplesDNS project, Joel De Gan added some comments to my De Gan Filters, Bloom Filters, Peoples DNS, and CNS post, and also asked What's the problem with FOAF? on the PeoplesDNS blog (also posted with additional comments by Marc Canter). Danny Ayers responded to Joel in Responding to Joel on FOAF.

Based on my work developing CNS for the iCite net, I wanted to suggest a way to think about how a PeoplesDNS (pDNS) based on FOAF might function:

level 0. pDNS root servers list each other (e.g., Root0 links to Root1 and vice versa, etc.). If you want / need to have more than a handful of pDNS root servers, there can also be a hierarchy of roots.

level 1. pDNS root servers list web pages (possibly some specific XHTML / XML / RDF format) that link to FOAF files (e.g., Root0 links to page1, page2, page14). These web pages effectively act like second-level DNS.

level 2. each web page that links to FOAF files also has a single link indicating the pDNS root server that lists the page (i.e., page1 is authoritatively listed by Root0, page22 is authoritatively listed by Root10).

level 3. each FOAF file links (back) to the web page that authoritatively lists it.

So, this assumes that each person has a single FOAF file, or that, if they maintain multiple FOAF files under the same "identity", they link them together within the FOAF.

With CNS, the one other notable feature is that the iCite containers (which are the equivalent to the level 2 in the above equation) can cache references to each other, thereby bypassing the need to use the root CNS server for each reference of a peer. The CNS server (level 0 and 1) is there to help account for when the level 2 information moves to different locations.

(There are a few other things about CNS that are added to this in terms of automatically redirecting / reconnecting things when they move, that functions across all four levels.)