DNS vs. XDI

Session Topic: DNS vs XDI: Who is better at solving which problems? What are the pros and cons in which situations?

Wednesday 2B and 5C

Convener: Esther Makaay

Notes-taker(s): Esther Makaay

Aspects for comparison:

• Discovery of attributes associated to names (+requirements)
• Registration (+characteristics)
• Trust of authenticity of discovered data (technical)
• Dependencies (decentralisation)
• Use/functionality: finding "what"/"where"
• (Suitability) (not in this phase)

Start of comparison based on explaining XDI by Drummond:

XDI

• Graph model (tree with references)
• Protocol (under development, OASIS)
  • Objects=graphs, sentences/ontology
  • Transactions: modifications (add, get, mod, del)
• Typed name structure (persistency)

DNS

• Tree model (no references)
• Protocol (RFC 1035)
  • Objects=RRs (separate protocols and RFCs)
  • Query/response protocol (modifications of RRs out of band)
• Typeless name structure

But how do these names co-exist?

• My subdomain: esther.makaay.nl
• My iname/cloudname: =esthermakaay
• Or should we fathom something like xdi:esther.makaay.nl -> RR …

(and then time ran out)

Session continued 3:30 PM

Let’s do a short paper on this topic, in two parts:

1) core capabilities / characteristics (Including objects exchange mechanisms)

2) scenarios / use cases ("How would you do...in...")

Examples and aspects mentioned to go into this:

• Data portability, Portable permissions / authorisation for a part of the graph, Performance, Revocation, Permutable data structure, RDF graphs vs RRs, XDI variables vs NAPTR wildcard resolving, Different 'types' of RRs: visible (usage/referencing) & administrative, Graph signing (and XML variants to do so) vs DNSSEC

Working lunch & post-closing time on Thursday will be used to flesh out this paper.